English - Compute Canada

smilinggnawboneInternet και Εφαρμογές Web

4 Δεκ 2013 (πριν από 3 χρόνια και 6 μήνες)

121 εμφανίσεις


1

|
P a g e







STRATEGIC PLAN


2010
-
20
20








HIGH PERFORMANCE COMPUTING:




BRINGING WEALTH AND HEALTH TO CANADIANS


2

|
P a g e


_____________________________________________________________________________________

Tab
le of Contents


___________________________________________________________________


1. Executive Summary

................................
................................
................................
......................

4

2. Authorization

................................
................................
................................
...............................

6

3. Vision and Mission


................................
................................
................................
.......................

7

3.1
Vision

................................
................................
................................
................................
.................

7

3.2

Mission

................................
................................
................................
................................
..........

7

4. Strategic Goal

................................
................................
................................
................................

7

4.1 Strategies to Achieve the Go
al

................................
................................
................................
............

7

4.1.1
Clarify the Governance Model

................................
................................
................................
.....

7


4.1.2
Bring
the
H
P
C
/
GDP Ratio

to the Mean for Industrialized Countries

................................
........

8


4.1.3
Provide Up
-
To
-
Date HPC Facilities

................................
................................
.........................

8


4.1.4
Develop
Highly Qualified Personnel

................................
................................
......................

9

5
. The
Value Proposition for
Stakehol
ders

................................
................................
.........................

9



5
.
1

The
Stakeholders

................................
................................
................................
..............................

9


5
.2 The Value Proposition

................................
................................
................................
.................

1
0

6
. The Current Context

................................
................................
................................
...................

1
2

7
. Organizational
History…………………………………………………………………………………………………………………
1
3

8
. International Trends

................................
................................
................................
...................

1
4

9
. Strategic Issues

................................
................................
................................
............................

1
6

9

.1 HPC Resources that meet the

needs of Canadian Researchers

................................
......................

1
7

9
.1.1
Desktop
/Small Lab

Computing

................................
................................
................................
..

1
9



9
.1.2 Mid
-
Level

Computing

................................
................................
................................
..............

19


9
.1.3 Tier 1

................................
................................
................................
................................
.....

20


9
.1.4 Highly Qualified Personnel

................................
................................
................................
..

2
1



9
.1.5

Technology and Timelines

................................
................................
................................
.

2
1


9
.2 Data, Data, Data

................................
................................
................................
.............................

2
2


9
.
3

Progra
ms and Outreach

................................
................................
................................
.................

2
3

10. Long
-
Range Budget Forecast

................................
................................
................................
......

2
4

10.1 Consequences of Underfunding

................................
................................
................................
......

2
4

1
1
. Performance Assessment Framework

................................
................................
........................

2
5

11.1 Objectives

................................
................................
................................
................................
......

2
5


1
1.1.1 National HPC Platform

................................
................................
................................
.....

2
5


11.1.2 Support for Research
ers

................................
................................
...............................

2
5

12. Conclusion

................................
................................
................................
................................
.

28



3

|
P a g e




APPENDICES

APPENDIX 1:
International Review Panel Recommendations

................................
................................
....

29

APPENDIX 2:
What Researcher Stakeholders Are Saying

................................
................................
..........

3
2

APPENDIX 3:
Stakeholder Comments on Digital Economy Consultation Submission

...............................

3
7

APPENDI
X 4:
HPC/GDP

................................
................................
................................
..............................

39

APPENDIX 5:
Strength
s, Weaknesses, Opportunities and Threats

................................
............................

4
0

APPENDI
X 6:
S&T Priorities Supported by Compute Canada

................................
................................
....

4
2




4

|
P a g e


1.

Executive Summary

High Performance Computing (
HPC
)

is
redefining the way that

research is done
. A
s a consequence,
research in all disciplines is d
elivering new knowledge and

innovation
, resulting in wealth creation, social
advantage and well
-
being for all Canadians.

Compute Canada’s
strategic plan is designed to
build on the
significant successes achieved to date and to
address th
e

opportunities an
d
challenges

that lie ahead.


Compute Canada’s HPC infrastructure
provide
s

a
national

platform

that

enable
s

Canadian researchers to
compete on an international
scale
,
attracts

top talent to Canadian

universities and broaden
s

the scope
of research
.

T
his na
tional platform has become an integral part of Canada's digital economy

as
demonstrated by t
he rapidly growing use

of Compute Canada’s hardware, software and people

resources

a
nd

evidenced by a growth of 30% in the number of researchers
supported
in the pa
st year
alone
; over 4,200 research
ers now rely on this critical infrastructure.


High Performance Computing
and High Performance Networking

are critical components
of digital
infrastructure

that also includes

repositories of complex data sets, a wide range of network
-
accessible
research equipment, digital devices and di
stributed sensors
, and

the related tools and services
.



Compute Canada and CANARIE are engaged in developing a strategic vision for digital infrastructure
(also known as cyberinfrastructure) in Canada. This vision will demonstrate the
fundamental
value and
benefits

of High Performance Computing and High Performance Networking to Canada’s economic
growth, achieve
cost efficiencies

in the service of academic and private sector research, and advance the
medical research necessary to secure the health and well
-
being of

all Canadians. We will discuss

integrating our infrastructures, our programs and our skills to deliver a world
-
class research
environment

and making the organizational changes necessary to bring a bold vision to fruition
.



Without a national platform HPC

resources w
ould

be isolated at
the
individual institutions that can
afford them. Canadian researchers w
ould

lose the ability to undertake a broad range of scientific
research and w
ould

lose the opportunity to work on science’s greatest challenges and coll
aborate on
international grand challenges.
Canada's ability to do much of the cutting
-
edge research
that

drives the
digital age

w
ould

be sacrificed
.


To ensure a sustainable national platform,
Compute Canada will r
ationalize

the number of
high
performance

computing cent
res

in Canada

and merge several existing consortia into regional divisions
.
We will take advantage of economies of scale while ensuring that the local support needs of the
Canadian research community are met. Th
is rationalized HPC landscape
will
make room for

a

new

Canadian Tier
-
1

cent
re that

is competitive on the world stage

and that forms part of a coherent HPC
ecosystem. This will ensure that Canadian researchers can leverage the capabilities of a rapidly
advancing technology
.


With

a ser
ies of other

concrete proposals on infrastructure, management and the evolution of user
support, this plan
addresses the recommendations made by the Mid
-
Term International Review
Panel
(
see Appendix
1

for details

and

www.computecanada.org


for
the full report

) and
identifies a clear way
forward for Canadian HPC.

It is

designed to improve service to our stakeholders and
to
broaden our user

5

|
P a g e


community by reaching out to non
-
traditional research segments
,

including

the humanities, medicine
and business interests.




GOALS

IMPLEMENTATION STRATEGIES

Implement and maintain a national platform
that contributes to and supports the very best
science that depends on HPC

Incorporate and staff Compu
te Canada; establish 4 regional divisions and
consolidate data centres; broaden the membership of the Board of
Directors
;

with CANARIE define a
strategic

vision of
cyberinfrastructure
/digital infrastructure

for Canada

and the organizational
changes require
d to implement that vision

Establish a Tier 1 facility and map the path to
exascale computing over the next decade

Bring the Canadian HPC/GDP ratio to the mean for industrialized countries
within a five year period by aggressively seeking additional fund
ing from
traditional and non
-
traditional sources and developing a plan to achieve
exascale computing in a timeframe that meets the needs of Canadian
researchers

Support HPC equipment and researchers using
HPC and train HQP
1

to harness HPC to perform
world

class research and improve
competitiveness

Work with universities to develop HPC support expertise and train
researchers to use HPC effectively and efficiently

Undertake HPC research to ensure that both
software and scientific applications are ready
for

advances in computer hardware


Build on the broad base of current Compute Canada users, extended to
Canadian participation in the G8 Funding Agencies exascale initiative and
work with NSERC, SSHRC and CIHR to develop an on
-
going program for HPC
professi
onals

Serve Canada’s existing and planned Canadian
and international major science
initiatives

Formalize relationships with the Canadian Astronomy Data Centre, SNOLAB,
the Canadian Light Source (CLS), NEPTUNE/VENUS, TRIUMF, ATLAS, CBRAIN,
Genome Canada,
the Square Kilometre Array (SKA), and other large science
facilities and ensure resources are available to meet their needs

Establish virtual HPC centres of excellence
initially for (i) SMEs (ii) medical research and (iii)
humanities and social sciences.

Establish HPC internships in
Small and Medium
-
sized Enterprises (
SMEs
)

in
conjunction with MITACS and encourage SMEs to work with academic
researchers; develop a pilot project with the medical research community;
develop a strategic accelerator plan, inclu
ding workshops and seminars for
the humanities and social sciences.









1

Highly Qualified Personnel (HQP): such personnel must be maintained and developed both as staff within
Compute Canada and users within the research community.


6

|
P a g e


2.

Board of Directors’
Authorization

This document was approved by the Board of Directors on December 21, 2010.

The Compute Canada Board of Directors:

Andrew Woodsworth, Chair





Ted Hewitt
,

V
ice
-
President (Research &









International Relations)
, University of









Western Ontario

John Hepburn,

Vice
-
President Research,



R.
Paul Young
, Vice
-
President,
University of British Columbia





Research,
University of Toronto

Jo
seph Hubert
,

Vice
-
rector, à la recherché,



Steven Liss
, Vice
-
President

Université de Montréal






Queen’s University


Christopher Loomis,

Vice
-
President, Research



Lorne Babiuk, Vice
-
President

Memorial University of Newfoundland




(
Research)

Universi
ty of Alberta

Rosie Goldstein
, Vice
-
President Research and



Alan Evans


International Relations, McGill University



Researcher Representative


Jonathan Schaeffer






Susan Baldwin, Executive Director

Researcher Representative





Compute Canada


Hugh Co
uchman, Chair National





Lynda Brulotte

Initiatives Committee






Chair National Administration










Committee



7

|
P a g e



3
.

Vision

and

Mission


3.1

VISION:
To

a
dvanc
e

research,

s
upport
and

accelerat
e innovation and excellence
, develop

highly
qualified
per
sonnel
,

and enable

competitive advantage,
economic prosperity
and well
-
being for all
Canadians

through the effective use of high performance computing
.


3.2

MISSION:
To create a
world
-
class
sustained
national platform of shared
high performance
computing a
nd data resources

and
personnel, accessible by
researchers in all disciplines
independent of
resource or researcher location
and to
promote high performance computing nationally and
internationally.



4.

Strategi
c Goal


Compute Canada will
contribute to a
nd
support world
-
leading scientific research

that depends on High
Performance Computing
, by


Implementing and maintaining a
world
-
class
national
HPC
platform
;


I
ncluding a Tier 1 facility and preparing for exascale computing;


Developing the Highly Qualified
Personnel required to support state
-
of
-
the
-
art HPC
equipment and world
-
class research using HPC and
t
raining Highly Qualified Personnel


capable of harnessing


state
-
of
-
the
-
art HPC equipment to perform world class research
and improve the competitiveness o
f Canadian Industry

;


Undertaking HPC research to ensure that both software and scientific applications
can
adapt to and take advantage of

advances in computer hardware;


Serving existing and planned
Canadian and international
major science
initiative
s;



E
stablishing virtual HPC centres of excellence initially for (i) SMEs (ii) medical research
and (iii) humanities and social sciences
; and


Developing a strategic communications plan to ensure that Canada’s accomplishments
in HPC
-
driven research and innovatio
n are better appreciated by all of our stakeholders.


4
.1

Strategies
to achieve the Goal



4.1.1

Clarify the governance model and staff Compute Canada appropriately


Incorporate Compute Canada and become membership
-
based at the level of
the institution (uni
versity, government labs, provincial governments, private
sector);


Focus on building a community
-
based organization rather than historic
consortia boundaries and
consider
the e
stablish
ment of

four regional divisions:
Compute West, Compute Ontario, Calcul Q
uébec and Compute Atlantic;


Re
-
examine the composition of the Board of Directors. For example, voting
Board members to include VPRs as designated by the regional divisions, 2
Compute Canada researchers, 2 representatives of government research labs, 2

8

|
P a g e


prov
incial representatives, 2 private sector representatives (1 vendor,1 user) ,
chair of the International Advisory Panel;


Regional VPR Board members must sit on the regional Board or management
committee to ensure national decisions are implemented locally;


Develop the management team, strengthen the role of the role of the Executive
Director with respect to operational decisions and add the staff necessary to
implement the defined strategies;


C
ollaborate with CANARIE and CUCCIO to define a joint vision of c
yber
-
infrastructure for Canada

and consider the implications for changes to both
organizations;


Consolidate data centres (machine rooms) while ensuring that support for
researchers remains locally available;


National rationalization of resources including

equipment, user support
personnel and funds. For example, new infrastructure funds, budgets for
operational funds will be prepared by each region and submitted to Compute
Canada for determination of the regional distribution; and


RFPs for equipment acquis
ition will be based on an equipment acquisition plan
and developed by a lead institution in conjunction with a national committee.
The RFP will be issued by an institution on behalf of Compute Canada.


4.1.2

Bring the Canadian HPC/GDP ratio to the mean fo
r industrialized countries within a five year
period by a
ggressively seek
ing

additional funding,
from
both traditional and non
-
traditional sources

in
order to serve a rapidly growing user base and accommodate technology advances



Request funding from CFI and
provincial governments to upgrade current
facilities and broaden the base of funding for both equipment and operations by
requesting funding from NSERC, SSHRC,
Industry Canada
and CIHR;


Seek funding for a Tier 1 facility
;



Undertake a gap analysis of the
skills required to support the research
community with respect to the three tiers of computation, the requirements to
prepare the HPC users for advanced applications as we approach exascale
capability, and
the
need for additional support personnel to fill ex
pertise gaps
such as data management, storage and archiving

and

software development for
exascale
; and


Examine the potential for a service fee structure as part of a sustainable
business model.



4.1.
3.

Provide researcher
s

with up
-
to
-
date facilities that a
re acquired and managed cost
-
effectively


Principles for determining the location of new equipment will include: support
expertise; projected costs for power, past performance, space, personnel,
renovation, etc.; contribution of institutions; agreement to a
dhere to
defined
service standards
;


Development of an equipment acquisition plan that targets the general purpose
clusters required by the majority of researchers, the equipment required by “big
science”
initiatives
, and the equipment
and support personne
l
that will be
required to
position Canadian researchers as world
-
leaders in their fields
; and


9

|
P a g e



Timelines for acquisition and implementation will be staggered over a 4 year
period in order to take advantage of advances in technology, economic value,
changin
g or unanticipated researcher needs,
and the
requirement to support
new “big science”
initiatives
.



4.
1.4

Develop
the
Highly Qualified Personnel

necessary for success


Decouple funding for personnel from funding for equipment to have the ability
to a
pportion funds as necessary;


Fill the gaps in service support (i.e. for non
-
traditional users, data management,
software specialists, new areas of economic benefit, medical research);


Train researchers to develop efficient applications and utilization of H
PC
resources, including those in the non
-
traditional disciplines and the private
sector;


Train support staff to support researchers and in particular to optimize
software;


Train staff to offer new services and to provide the required change
and
development
of code for pet
ascale and exascale; and


Enter a MITACS
Accelerate

agreement to develop an HPC internship program.



5
.

The Value Proposition for Compute Canada Stakeholders

5
.1

The Stakeholders

Compute Canada’s stakeholders include academia, federal and provincial govern
ments, funding
agencies, the private sector,
government
-
funded science
initiatives
, national and provincial

high
-
speed
network
ing organizations

and the Canadian taxpayer
. Some of these stakeholders, including
astrophysicists
, the sub
-
atomic physics
and cli
mate science
communit
ies

and the humanities research
community, have
explicitly
referenced the need for high performance computing and
stated
the desire
to work with Compute Canada to ensure a dynamic HPC environment for Canadian researchers (see
Appendix
2
). Compute Canada’s submission to Industry Canada’s Digital
Economy Strategy

Consultation
received the most votes, demonstrating a broad base of support from stakeholders.
A representative
s
ampling
of comments
is

included in Appendix
3
.

In the past, the
traditional core areas of research
requiring HPC

facilities have been physics
,

chemistry

and engineering
. While the
y

still predominate, these areas
themselves now

includ
e

specializations such
as bio
-
chemistry and bio
-
physics

and reflect the continuing mult
i
-
disciplinary trend in research.
Compute Canada resources are being used to fight infectious bacteria, diagnose Alzheimer’s disease,
track ocean
circulation variability

and
predict future
climate
evolution
, map the “Big Bang”, develop
everything from fuel

efficient cars to safe plastic,
and
analyze markets for more
useful

risk indicators
.

World
-
class scientists are being attracted to do their research in Canada through the Canada Research
Chair Program and the Canada Excellence Research Program. A majority

of these researchers and the
teams of graduate students and post
-
doctoral fellows working with and learning from them, whatever
their
field
s,

require
HPC

facilities and support from
Compute Canada’s

highly qualified personnel.
Indeed, major science
initia
tives
, funded by the federal government, such as ATLAS, the
SNOLAB
,
VENUS/NEPTUNE, astronomical observatories and the Canadian Light Source, among others, require

10

|
P a g e


supercomputing and are relying on the use
and continuing enhancement
of Compute Canada facili
ties
for their research.

Computers have re
-
invented everything we do.
High Performance Computing is

re
-
inventing how
research is conducted and
is
accelerating the timeframe for that research. Some businesses are already
using HPC to advantage. The movie indu
stry uses HPC for animation and rendering of special effects;
retailers regularly use data mining; and credit card companies use HPC for fraud detection. Simulations
to design chemicals that help the immune system fight bacteria now take three months on on
e of
Compute Canada’s new
high performance
computers rather than the more than ten years it would have
taken without it. The Canadian academic
commu
nity

and Compute Canada are working with the private
sector through investment in R&D to realize such orde
r of magnitude impacts on the
efficiency of the
Canadian economy.

The Canadian taxpayer is a direct beneficiary of the research that requires HPC through improvements
in the diagnosis, prevention and cure of disease; better understanding of pandemics so th
e general
population can be better protected;

green energy; understanding the impact of climate change and
defining how to stabilize and improve our environment;
food quality and safety throughout the food
chain; and almost every aspect of health and well
-
being.

5
.2

The Value Proposition


Compute Canada presents a value proposition that is central to economic prosperity, to an international
reputation for leadership in science and research, and to the social and cultural lives of Canadians.


High Performanc
e
Computing brings

economic wealth and health and well
-
being to Canada and

Canadians by
gaining knowledge
through the rapid solution of complex problems. The scope and
complexity of these problems may be hundreds or thousands of times

greater than could
be tackled by a
high end PC or could be solved a thousand times more quickly. Within the decade there is the potential
that thi
s advantage could reach one million times.


As stated by the G8 Research Councils Initiative on
Multilateral Resear
ch Funding “Major challenges of the 21
st

century, such as climate change, energy,
water, environment or natural disasters, can be addressed by high performance numerical and symbolic
simulations that are both data and computer intensive.”



Examples of th
e

value
proposition, through research currently being conducted using Compute Canada
resources i
nclude:



Finding genetic variants associated with specific diseases and developing drugs
to target them;


Understanding the basic forces that shaped our universe
and will determine its
future;


Researching dementia diseases leading to prevention, diagnosis and
management
. This

will significantly less
e
n the healthcare burden and economic
consequences which are projected to be $872 billion from 2008
-
2038

and by
2038
257,800 new cases per year
;

2.8%
of the population
; and

$153 billion
annually

(
“Rising Tide: Impact of Dementia on Canadian Society; Alzheimer
Society 2009);


11

|
P a g e



Increasing the productivity of crops, ensuring their nutritional value and feeding
Canada and other coun
tries;


Researching diabetes for prevention and cure (“The direct cost of diabetes now
accounts for about 3.5% of public healthcare spending in Canada.”

An Economic
Tsunami: the cost of diabetes in Canada, 2009
);

and


Identifying and understanding the effect
s of climate change, including
forecasting water resource and sea
-
level rise impacts from glacier retreat and
greenhouse induced global warming.


More generally, t
he value proposition
is established by
:


Support for excellence in scientific research that requires HPC equipment

and
expertise;


Acceleration of research and innovation through the use of advanced HPC that
enhances collaboration both nationally and internationally;


Certainty that major science initiatives can be assured of the availability of HPC
resources;


Inclusion

of medical research, humanities, social science and arts and design
within the support system;


Increasing the return on investment for federal and provincial investments in
R&D;


O
utreach activities and programs

such as HPC skills training;


An HPC interns
hip program in coordination with MITACS
2
;


Collaborative research between academia and Small and Medium
-
sized
Enterprises;


A cost
-
recovery program for SME access to HPC equipment and consulting
services;


Consultation with the research community to determin
e the need and timing
for data storage and archiving;


Increased private sector investment in R&D;


Access to resources across the country, independent of the location of the
researcher;


Naturally forming centres of expertise in manufacturing, development of

green
technologies, new products, emerging industries;


New world
-
class initiatives with Canadian leadership; and


Repatriation of world
-
class scientists.


Canada has a productivity gap relative to other G8 countries. The role of
technology
generally, and h
igh

performance computing in particular,

in realizing a reduction in that gap cannot be overstated
. High
performance computing is one of the necessary technologies that will contribute to that reduction and
an improvement in
innovation, productivity

and e
conomic performance.
High Performance Computing is
essential to creating new knowledge and generating the knowledge that results in innovation
.

It is
innovation that raises domestic productivity levels which in turn achieves economic growth.

High
Performa
nce Computing and High Performance Networking are the fundamental tools of innovation.




2

MITACS is an organization that brings together academia, industry and the public sector through research and
training initiatives to develop cutting edge tools vital to the knowledge
-
based economy.


12

|
P a g e




Supercomputing is part of the corporate


arsenal to beat rivals by staying one step


ahead of the innovation curve.
(
The New Secret Weapon”,

the Council on Competitiv
eness, 2008
)








“Productivity growth is the most








important measure and determinant







of a nation’s economic
p
erformance.”








(S. Ezell & R.D. Atkinson,
The Good, the Bad and the







Ugly of Innovation Policy, October 2010
)

Ca
nada has been out
-
stripped by our competitors in both High Performance Computing and High
Performance Networking. It is imperative that we “make up the distance”

if we are to have a future in
the global marketplace and a population confident that their hea
lth and
social and economic
well
-
being
are

secure.


6
.

The Current Context

The r
esearch
enterprise
now includes computation as
a

third
distinct methodology,

the other two being
the traditional areas of laboratory experiment

and theoretical

analysis
.
Numerical s
imulation and
m
odeling advance
the
understanding of complex phenomena and contribut
e

to innovation in a wide
variety of sectors from automobile to aerospace to pharmacology to animation to disease prevention
and control.

Tasks that can take months or years using “normal”

computers can be
completed

in days or
even minutes with high performance computing
; these systems also allow investigations of a scope that
would otherwise be simply impossible
.


A component of the digital infrastructure that is critically important for h
igh performance computing is
networking. Canada’s national
advanced
network

for research, industry and education
, CANARIE, and
the
regional ORANs

must continue to evolve and be capable of meeting the

technological requirements
of advanced computing and th
e

corresponding
expanding
needs of the HPC community
.

Computing
power is increasing exponentially
. D
ata management, storage and accessibility requirements are
expanding
. T
he potential weak link is networking.

As we implement petascale
computing
(
the curr
ent
Tier 1

level
) and plan for exascale

computing
, it is imperative that Canada’s national network focus on
the continuing development of

an advanced network that can
deliver the

required

capabilit
y
. There are
also
a number of “last mile” issues that must
be addressed by
the provincial Optical Regional Advanced
Networks (
ORANs
) and

universities
within provincial boundaries in order that researchers can truly take
advantage of the national shared computing platform and collaborate nationally and internationa
lly in
the knowledge and scientific challenges that lie ahead. Compute Canada will continue to work with the
network providers
at the national, provincial and campus levels
to ensure that networking
is an
accelerating factor for research and
does not beco
me

an inhibiting factor.

A
common
digital
infrastructure
vision

will
help
ensure that the critical components are mutually reinforcing and provide
research, economic and social advantages.



13

|
P a g e


As the world generates ever more data and as more and more data t
hat was generated in the past is
digitized, the
only
way to understand and
extract its value

is with high performance computing.

7
.

Organizational
History


Seven regional HPC consortia were formed be
ginning in

1998

in response to the creation of CFI and i
ts
first call for proposals
.
In 2005,
the
C3.ca Association Inc.
, an organization that brought coherence and
visibility to high performance computing and supported the development of the regional consortia,

published its Long Range Plan for High Performan
ce Computing in Canada:
Engines of Discovery: The 21
st

Century Revolution.
In 2006, the Canada Foundation for Innovation created the National Platform Fund
program in part as a response to the Long Range Plan

(LRP)
. In 2007, the Long Range Plan for HPC in
Canada was re
-
examined in light of the significant accomplishments of the past decade and the promise
of the future.

The Long Range Plan remains a valuable position paper that highlights the use, benefits
and contributions of HPC in Canada. It defined the
HPC community and united it with a common vision
and sense of significance. It led to the establishment of Compute Canada as a national
platform and a
national representative

for HPC.



This strategic plan builds upon the
LRP
and presents objectives and st
rategies for the implementation of
the vision it presented.

The LRP recognized that Canada has established a strong international position
with respect to mid
-
range HPC facilities and that, as a result, many outstanding researchers have been
attracted to C
anadian universities and helped to grow Canadian industry. It stated:


“If we are to keep these people and gain the full benefits of the investment,


this HPC infrastructure must be sustained at a competitive level. If we are also



to address the grand
-
c
hallenge problems, it will also be necessary to establish



a high
-
end computing facility. … With

this investment [
$76M in 2006, $87M in


2009 and $97M in 2012
]

and these facilities Canadian researchers can be leaders



in discovery and innovation,
and
Ca
nadian

industry can be internationally


competitive in a new and lucrative

range of fields.”


Since its inception, Compute Canada has made decisions in the national interest of all Canadian
researchers
while providing

the flexibility to ensure the most ef
fective and efficient implementation of
these decisions at the regional level.

The creation of a Compute Canada Data Base has ensured a single
point of entry to Compute Canada and its constituent parts. A National Resource Allocation Committee
has been est
ablished and
two national

calls for proposals issued
. There is a single point of submission for
all allocation requests regardless of where the researcher is located or the machine requested and there
is effective coordination between the national and regi
onal levels

to ensure optimum service and value
to researchers
. In terms of interacting with and understanding the needs of the researcher community,
town hall meetings were held across the country thereby providing input to this strategic plan; a hands
-
on

workshop for humanities researchers
assisted proposed

projects
to get
up and running and
effectively using HPC resources; and an HPC for health research
workshop
was held

leading to
discussions centered on the utilization of shared resources. Compute Cana
da provided strong support to
NSERC to ensure Canadian researchers could participate in the G8 funding agencies exascale
competition. Compute Canada and MITACS have signed a Memorandum of Understanding that will lead
to establishment of HPC internships, pr
oviding advantages to graduate and post
-
graduate researchers as
well as to Canada’s business community. And the Board of Directors has expanded to include two
researcher representatives.


Compute Canada continues to gain visibility nationally and
is
inter
nationally

recognized as representing Canadian HPC
.


14

|
P a g e



The community of researchers
that

require
s

HPC is growing rapidly and
Compute
Canada must

ensure
that
the

HPC resources
it provides
track the research needs. Demand arises both from the growing user
base

and from the requirements of researchers to remain competitive as the technology advances and
system sizes and capabilities increase.

As

new machines
become available
,
experience has shown that
the additional capacity
will

be fully utilized
almost immedi
ately
. The existing
demand is rapidly
outstripping the current
supply
.


In 2008
-
2009,
the number of

researchers us
ing

Compute Canada resources

grew by

almost 1000 or

30%

over the previous year
. With the addition of new machines in early 2010 and early 2011
, more
researchers with increasingly complex applications are expected

to
request access to
Compute Canada
resources
. There are
also
researchers in Canada at present whose work would be significantly advanced
if Compute Canada had a
Tier 1

system in place
.



Compute Canada is supporting Canada’s major science initiatives, including ATLAS, NEPTUNE,
Ocean
Networks, the Canadian Light Source,
SNOLAB
, and astronomical observatories, among others.

Compute

Canada also supports

private sector initiatives such as
the Cancer Biomarker Network.


The
HPC
platform

and the highly qualified personnel who support it are an integral and essential
component of Canad
ian research infrastructure.



8
.

International Trends

Canada, as part of the global community,
faces many r
esearch challenges such as those related
to
climate
c
hange, pandemics, and energy and water shortages. Canadian researchers depend upon the
availability of competitive high performance computing facilities in order to participate in addressing
these “grand

challenges” as well as
collaborating with other researchers in Canada and around the
world on these and
many other economic, social, and medical challenges
.



On a global basis, the
re is a clear

trend

of

increas
ing

investment in high performance computing

by
both
national
and
state governments
, and by the European Union

(EU)
.
According to the IDC Special Study,
July 7, 2010, (
D2 Interim Report:

Development of a Supercomputing Strategy in Europe): “the
supercomputer segment is in a high
-
growth mode, even wi
th the current recession


it grew 25% in
2009.”

Appendix
4

shows the HPC investment
relative to

GDP for 40 countries. Canada ranks 24
th
.

And it
is not just the advanced or developed nations that are investing. Cyprus, Malaysia, and Bulgaria are all
investing mor
e
in
HPC/GDP than Canada

is
.

Even in
such
difficult financial times, countries such as Australia, New Zealand, U.K. (Wales) are making
significant investments in high performance computing infrastructure and personnel. There is a common
understanding among

the most productive nations that HPC is
a fundamental pre
-
requisite for

economic
security and
performance, competitive advantage and productivity growth.



Old technology cannot compete with new technology


and the field of HPC move
s

very quickly.
Resear
chers using old technology are at a competitive disadvantage with the result that the innovations
that might result from that research cannot keep pace with the number and speed of innovations
by
researchers
in other countries.
Canada, through the Canada
Excellence Research
Chairs
Program (CERC)
program, has been very successful in attracting the best and the brightest researchers from other
countries.

It is imperative that we provide the best infrastructure to support their work and reap the

15

|
P a g e


resulting ben
efits in innovation, excellence, advantage and reputation that these researchers
are

expected to bring to Canada.

One of the recommendations in the
IDC Recommendations Report: For EU HPC Leadership In 2020 states
“It is recommended that the EU and the nati
ons make HPC a higher priority and step up to either the
“Full leadership level” or at least the “Funding to reach major goals level” scenario level: This would
require a
net new investment reaching 600 million euros a year within five years”
.


According t
o the Oak Ridge National Laboratory the predicted
performance of

leadership computing
platforms
will grow from

under a petaflop in 2005,
and

one petaflop in 2010,
to

20 petaflops in
approximately 2012, 100 p
etaflops by 2015 and 1 exaflop (1000 petaflops)
i
n approximately 2018.
Japan
is targeting 10 petaflops for 2011.


PRACE (
Partnership for Advanced Computing in Europe) is a pan
-
European research infrastructure for
high performance computing

that is funded by partner countries and the European Union’s 7
th

Framework Programme. PRACE “
provides Europe with world
-
class systems for world
-
class science and
strengthens Europe’s scientific and industrial competitiveness.”

PRACE will install
its second

Tier
-
0
[defined as one of the few largest systems in the world
]

system

in France before the end of 2010 and
three more Tier
-
0 systems
w
ill be deployed by Italy, Spain, and Germany. Each of those countries has
committed 100
m
illion
e
uros for PRACE resources

and remain
s

committed
to supercomputing
in spite of
serious economic difficulties
. All the related procurements will be made by the end of 2013.
Each system
will provide several petaflops of computing power. PRACE is targeting exaflop computing power

by
2019
.


In
June 2010 the Governor of Massachusetts
(
population approximately

6.5 mi
llion)
announced a project
to “create a world
-
class high performance computing center that will provide an infrastructure for
research computing in life sciences, clean energy and green computing; help establish a collaborative
research agenda; and catalyz
e the development of an innovation district in downtown Holyoke.”


Wales
(
population approximately

3 million)
is investing $60

million in regional HPC capability (includ
ing

equipment, distribution networks,
and training
)
.


Cutting
-
edge computing facilities wi
ll be available for
use by businesses working independently or in collaboration with academics and will establish Wales as
a key international centre for specialist computational research.” (Wales
, Assembly

Government, July
2010)


Exascale computing is on
the horizon and initiatives are underway

to examine the software and
middleware needs.
NSERC sought the advice of Compute Canada
regarding

whether or not to
participate in
the

initiative by the G
-
8 funding agencies “Interdisciplinary Programme on Applicati
on
Software toward Exascale Computing for Global Scale Issues”. Compute Canada
strongly
supported
Canadian participation. This is
critical in order to

position

both

Canadian researchers
and the analysts
who support them
for the future of high performance c
omputing

and, therefore, the future of research
.
That being said, Compute Canada should be providing a
current
Tier 1

facility now as there are
significant numbers of researchers who
need
that capability
to be competitive
as well as to ensure that
those re
searchers who will require exascale capability
when it becomes available
will be
part of a
community that has the skills to be
able to use it
.




16

|
P a g e


T
he software and methodologies
required
in order to ensure that researchers can use
an exascale

machine are sig
nificantly different
from

those used today.
This, too, represents
a
significant
research
challenge.
It is essential t
o ensure that Canadian
academic
researchers


and by extension the private
sector
-

are positioned to take advantage of such technologica
l advances
.

I
t is imperative that
Compute
Canada

work with international teams to develop new approaches

and prepare both academic and
private sector researchers for early adoption of this pivotal technology
.
By investing

in the ability to
leverage HPC acr
oss the spectrum of capability
,

Canada’s

private sector will be
well positioned
to
compete in the

digital
econom
y



an economy

that will rely
heavily
on

high performance c
omputing.


Compute Canada must be able to position its research community to be part
of an international
community of research using state
-
of
-
the
-
art HPC facilities. While we do not need to duplicate the
models of other countries, we need
a
credible presence
to
allow our researchers to
be fully engaged
participants in

the global
HPC

commun
ity and, therefore, plausible partners in joint undertakings.

At
present
,

i
nvestments in HPC in Canada are between 35%
-
50% of the levels
of

our G8 co
unterparts
.

Canada’s partners in the G8, which have experienced significantly more severe reversals of their

economic fortunes than has Canada, appear to be stepping up their commitment to HPC rather than
diminishing it
.

This is a clear indication that
high performance computing

is understood to provide
an
important impetus to innovation

and, hence,
is
a key com
ponent of recovery and growth
.


9
.

Strategic Issues



The identification of the strategic issues is the result of
the analysis of
Compute Canada’s strengths,
weaknesses, opportunities and threats

(SWOT)
.



Compute Canada’s strengths centre on the policies,

mechanisms and operations that are driving the
implementation of a national HPC platform and its consolidation and integration
, on

ensuring
the
availability of
substantial infrastructure and
on enabling
researcher access to machines and expertise
across the country.


Compute Canada’s w
eaknesses relate to the need for a robust and sustainable business model and a
strengthened governance structure. As stated in the Mid
-
Term Review submitted to CFI: “Compute
Canada’s current financial model is arguably its greatest weakness and poses a sig
nificant challenge
going forward”.


Compute Canada’s opportunities include outreach to many different research communities including
the private sector in order to extend the use of HPC for innovation and economic development and
ensuring we are prepared w
ith the people and the expertise to assist them.


The threats facing Compute Canada come from a variety of sources ranging from the lack of recognition
generally of the strategic importance of HPC to research, innovation and the economy
,

to the increasing
c
ost of power
,

to not knowing if or when new funding for HPC
resources
will become available.


The complete SWOT listing may be found in Appendix
5
.






17

|
P a g e


9
.1

HPC

Resources

t
hat

meet the needs of Canadian researchers


Compute Canada is a national platform and as
such
must
support

researchers in many different
disciplines who have different requirements

and

varying skill levels with respect to HPC, and projects
that range
in scope
from an individual project to multi
-
country projects requir
ing

international
agreemen
ts that stipulate the HPC commitment.

The use of high
-
performance computing across many
research disciplines is evidence of its
necessity
as a scientific tool
.

A national platform is composed of
highly qualified personnel,
facilities,
and software.


Discip
line
-
specific facilities such as a telescope, a particle accelerator or a neutrino observatory become
rallying points for their communities of interest. It is much harder for infrastructure platforms that serve
many disciplines to create that same sense of

community. As an example, there was little interest in the
internet among the Canadian research community twenty years ago. Now, the internet, like high
performance computing, is fully acknowledged as an indispensable research tool.



Compute Canada
and C
ANARIE are

key
component
s

of
Canadian
digital
research infrastructure
.

The
national HPC
platform
,
w
hich

is relied on

by
Canadian researchers and their international partners
,

can

be sustained

only

if
there
is

predictable

and sustained funding.


T
here are
several
types of user support

expertise and
infrastructure
require
d

within the Compute
Canada user community
.
Although t
he facilities and expertise required to support
the diversity

of
research

groups vary significantly
,

both user support personnel and the

computing systems themselves
may be effectively shared.


This sharing of resources is critical to their
efficient and effective use

as well
as to maximizing their value to
all stakeholders including
the research community and those who
provide funding for

those resources.


Compute Canada's computational capability and storage facilities have grown from a modest beginning
to the current offerings that make up the national platform
.
Total capability will expand slightly in the
first half of 2011 as the final

components of the
NPF
(
National Platform Fund, a Canada Foundation for
Innovation program)
are deployed. After

2011 the older, pre
-
NPF
(National Platform Fund, a Canada
Foundation for Innovation Program)
hardware will be decommissioned reducing Compute C
anada's
total capabilities
significantly.
Compute Canada will begin decommissioning the NPF funded hardware
in

2013 with all those facilities be
coming obsolete b
y 2016.




18

|
P a g e



Figure
1

: The Computi
ng Ecosystem



In order to serve its broad base of constituencies

and to provide pathways for researchers to develop
expertise to use more sophisticated systems as demanded by their
research
programs
, Compute Canada
must provide a full ecosystem of capabil
ities and expertise.
High Performance Computing
requires the
availability of

many different capabilities from desktop
/small lab

computing, to mid
-
range computing to
the very largest systems as well as
new
architectures

such as
GPGPU

and cloud computing.
In

Figure 1
we note that the collectivity of Compute Canada
’s

mid
-
range capability may be considered an incipient
“cloud”.



Computing is itself a tool that continues to evolve. Compute Canada will continue to support all levels of
capability
from the mid
-
ra
nge to Tier 1
and

will

introduce the

required

new
architectures

for the benefit
of the research community.
This
support
encompasses not only the

machines

but also

the

infrastructure
and
user support staff
necessary
to meet the needs of
both
individual rese
archers and research groups

at the national and international levels
.

The Tier 1 and mid
-
range levels would include both user and
infrastructure support. At the desktop
/small lab
computing level, Compute Canada provides support to
researchers

primarily in
the form of advice to enable them to move into the higher levels of the
ecosystem

in order to expand the objectives and capabilities of their research.


A computing ecosystem must be

built for performance
and
must
optimize

the range of capabilities and
cap
acities
of the

research community

it serves
.


The computing ecosystem proposed by Compute
Canada has three levels
as identified in

F
igure 1.


CFI funds a portion (approximately 40%) of
capital

and some part of the operational needs; however
, in

order to
meet the needs of researchers capable of developing a strong HPC capacity and capability for
Canada
, Compute Canada

needs to offer much more than simple access to infrastructure
. Compute
Canada must design and offer programs that enable researchers to

make

effective use of
the
infrastructure

and to
assist and collaborate with researchers

in order to elevate their research output

and
enable

them compete on the world stage.

We

must also engage in outreach to private sector SMEs
and to government science depa
rtments in collaboration with university
-
based scientists. This is the
opportunity that can be exploited through the provision of the full computing ecosystem.


19

|
P a g e



9.
1.1

Desktop
/Small Lab

Computing



The
basic

level
of the computing ecosystem
is
composed

of l
ocal computing facilities, typically desktop
or small lab
systems that are used to develop and test applications to be ported to the next level

and

to
perform modest visualization tasks and analyses
.

Acquiring and managing
such systems
is

the
responsibili
ty of

the
researchers themselves
.

T
his level
also

include
s

clusters that can be quite
substantial, although
typically far from

the Top500 category.


One of the most important services
Compute Canada offers to researchers at this level is user support.


To
advance their research or remain competitive in their fields the majority of these researchers will
need to
take advantage of

the mid
-
range level of the computing ecosystem

either for specific projects
or, increasingly, as their new mode of computing.

Comp
ute Canada works with researchers at the
desktop
/small lab

level to accelerate their progress to mid
-
range in support of excellence and
innovation.
Some researchers within the humanities and social science communities are currently
working at th
is

“staging
” level. These researchers have recognized the need to move to the mid
-
range
and have approached Compute Canada for assistance in making that transition. Compute Canada will
continue to offer hands
-
on workshops and seminars as well as one
-
on
-
one support to

enable this.



9
.1.2

Mid
-
Level

Computing


The second, or mid
-
range level, is
composed

of a variety of systems that typically
rank on

the
Top500
on
installation

and
that
are the main production nodes of the ecosystem.
Compute Canada’s mid
-
range
systems w
ill be consolidated to ensure a viable mid
-
range capability and will

be located at
approximately ten sites in Canada. These facilities are intended

to satisfy the needs of the majority of
researchers.
This category

constitutes the backbone of the HPC ecosy
stem.


Compute Canada has internationally competitive mid
-
range facilities that are supporting a wide variety
of researchers in very diverse fields. The majority of the researchers who use Compute Canada’s
facilities use mid
-
range equipment. These
resea
rchers are

undertaking significant work and making
significant contributions to the base of knowledge
across a wide range of disciplines.
These are the
researchers who are positioned to join international collaborative projects based on the world
-
class
exp
ertise they have
demonstrated

while working in the mid
-
range. These researchers are addressing
priorities such as environmental science and technologies, natural resources and energy, health and
related life sciences and technologies, and communications a
nd information technology
.



It is imperative that these facilities are maintained and renewed in order for our researchers to remain
competitive over time and to continue to contribute to the Canadian economy, culture and society and
the well
-
being of it
s citizens.


Mid
-
range systems

which will be

distributed geographically

in order to leverage existing data centres
and provincial contributions
, will be equally accessible to all
C
anadian researchers and
the mid
range

will
include

as a

component
cloud

computing
.

C
loud computing plays a valuable
role in enabling Compute
Canada

to support
the
diverse and non
-
traditional user communities
which are
part of its mandate.
Cloud
c
omputing testbeds are
currently

operating on Compute Canada hardware
in

support

of

20

|
P a g e


astronomy, medical

and other research computing. Compute Canada has
specia
lized

expertise in this
area and will continue to provide leadership in this rapidly
-
developing
computing
methodology
.


9
.1.3

Tier 1


The highest level
in the computing ecosystem

consists of

world class
Tier 1

facilit
ies
. A
t this time
, such a
facility wou
ld be a petascale

computer,
although this

definition evolves with the
rapid

progression of
computing power. A Tier 1 facility
w
ould
typically
be ranked in the top 20 facilities worldwide and
would
be

designed to
allow
very
large
-
scale, parallel computati
ons at the leading edge of research in many
fields.
Applications on such systems currently run on several thousand processing cores or more and this
requires special architectures,
especially

a fast interconnect

to be effective research tools. Given the
sp
eed of technolog
ical

advance, current Tier 1 systems will be mid level in three years. Given the
challenges in using
such

systems, a

Tier 1 system is necessary to prepare HPC users
to move to

the next
generation of mid
-
range
facilities:

as computer power i
ncreases, users worldwide
must

follow and adapt
to new technologies in order to be fully productive when new mid
-
range systems are installed
en
masse

.
Access to a Tier 1 facility provides researchers with a four to five year
advantage which will
dramatica
lly

increase their competitiveness.


As new frontiers of science develop in fields from the human genome project to biochemistry, physics
and material science, a large

scale computation facility is
a pre
-
requisite for that research. Such a facility
does n
ot exist in Canada and
Canadian researchers require a
Tier 1

facility now.
This could be achieved
by significantly upgrading an existing Compute Canada facility which
has
demonstrated the

capab
ility

of
managing high performance systems or it could be
a
n en
tirely

new
facility
.

T
ier 1

is an important step in
preparing
both
researchers
and support analysts
for the exascale equipment that is expected in the
2018 timeframe

and which is widely believed to present very significant challenges for the HPC
community
as well as unprecedented opportunities for new results and breakthroughs.


Appendix
6

provide
s examples

of

research objectives
that
can be met with existing capabilities and th
e
unprecedented new knowledge and advantages that could be achieved with Tier 1
capability.
Examples
are provided for each of the four strategic priorities identified in the Government’s Science and
Technology Strategy: environmental science and technologies, n
atural
r
esources and
e
nergy,
h
ealth and
r
elated
l
ife
s
ciences and
t
echnolog
ies, and
c
ommunications and
i
nformation
t
echnology.


There are economic and social advantages that result from the knowledge gained by participating in
research at this scale. The knowledge results in innovation within a wide variety of sectors. Canadian
r
esearchers must have the HPC facilities necessary to earn a place at the table
where

such challenges

are addressed
. A good example
is the Canadian team that assisted in designing, building and operating
the CERN facility in Switzerland and was involved in
the first set o
f results from the Large Hadron

Collider.

Without
HPC and

associated networking capabilities
“at home” those researchers could not
have participated

and Canada would not be in a position to benefit from and capitalize on the “spin
-
off”
know
ledge that is inevitable from an undertaking of this magnitude
.


HPC is dynamic and what is
Tier 1

today will be mid
-
range in 2
-
3 years and will be almost obsolete with
the arrival of exascale. Canada needs an internationally competitive Tier
-
1

facility an
d today that is
Petascale.
With a Tier 1 facility, Canada’s researchers will accelerate scientific discovery, enable
researchers to conduct transformational science, and
enable innovation throughout all levels of the
economy.


21

|
P a g e



9
.1.4


Highly Qualified Perso
nnel

The key “product” academia provides to businesses is Highly Qualified Personnel. The development of
highly qualified personnel constitutes the most effective form of knowledge transfer. Computationally
trained individuals are a
very valuable

output o
f the academic environment. These individuals go into
businesses and enable th
ose businesses

to take advantage of HPC for reducing
the time for innovation
and
the time to get a product to market. We must continue to develop these talented and creative
peop
le in order to position our
private sector

to be early ad
o
pters of
new
computing

power and
technologies

and realize the resulting economic advantages and wealth creation.


As computing architectures evolve to provide enhanced performance, the associated so
ftware, both
system software and applications software, must be re
-
written to accommodate the new features. This
highly specialized s
oftware
will continue
to

need to

be re
-
written and a new generation of software
tools
must be
developed in order to keep pa
ce with the
architectural advances.

This is both a challenge
and an opportunity for Canada.

Individual scientists and researchers are
usually
not specialists in the development of the software they
need
to perform their research.

Given the growing comple
xity of data sets and the exponential growth
in the volume of data available to researchers
,

it is imperative that we

increase the numbers of highly
qualified HPC personnel available to
support
the research community.
For this reason, HPC specialists
are an

essential component of
the
scientific research that is dependent upon HPC.
These individuals are
part of the competitive advantage we provide
. At present, we do not have sufficient numbers


in either
academia or business.

“The investments required for sc
i敮瑩晩挠慮T 散onom楣i獵捣敳猠楮捬cTe慮y
慲敡猠楮⁡摤楴ion⁴o⁴U攠cos琠of⁴U攠捯mpu瑥r献⁔桥smo獴s捲楴楣慬⁡牥a⁩猠
Uum慮⁥硰敲瑩獥H⁩湣 uT楮g⁴U攠獣s敮瑩獴e⁡湤⁲ s敡牣e敲猠慳a睥汬⁡猠sUe
數p敲瑳⁩渠 獩sg⁴U攠獵s敲eompuW敲献⁔U敲攠楳⁡ gro睩ng⁷o牬r睩w攠
獨s牴慧攠
o映䡐䌠C慬敮琠摵攠瑯⁡渠慧ing⁷o牫ro牣攠慮T⁡ 獣慲捩sy映湥眠g牡ru慴敳e楮
various HPC fields.” (IDC 2010)



9
.1.
5

Technology and Timelines


The rapid and continuous
progress in computer
technology is

a powerful driver of the HPC environment
and
community.
T
his dramatic growth enabl
es researchers to tackle

problem today that were
unimaginable a few years ago. It
h
as
also
resulted in a number of strategic challenges that Compute
Canada must meet i
n order to leverage

these trends effectively. This s
ection gives an overview of
two
key

challenges
: power consumption and software development
.


The scale of modern massively
-
parallel systems presents extreme challenges in terms of the
infrastructure required to support them, particularly power and cooling.

A
lthough the per
-
processor
power
requirement
is now largely static, the overall power consumption continues to rise as the number
of processors per system increases. Predictions for the first exascale system are that it will consume
1GW of power although
it is believed that real power consumption w
ill

have to be under 100MW for it
to be feasible.
To address this,
Compute Canada will
consider the cost of power as an important
criterion in determining the future locations of

data centres
.


22

|
P a g e



T
he combination of
the sheer scale of large systems together with their
increasingly
heterogeneous
nature makes writing efficient software for them an extreme challenge.
It is generally recognized
that
current

architectures have

far outstripped the availability of software t
hat can take advantage of
their
capabilities
.
As an example,
whe
n

an application program must take advantage not only of all cores in a
node but of thousands of cores across nodes, the challenge can be extreme
.

These trends highlight the
need for skilled t
echnical support, including programmers, to ensure that Canadian researchers are able
to take advantage of contemporary
and
future systems. It will be increasingly challenging for individual
programmers to develop applications for such systems and the poss
ibility of writing a one
-
off program
for a specific short
-
term need for such a machine is becoming remote. Compute Canada will actively
encourage and enable the use of these systems


at both the mid
-
range

and Tier 1
levels


through
training and programmi
ng support.
T
he development of a culture of software

is

a long term investment

and one in which we must begin to invest immediately.

These activities are labour intensive and will
require additional staff.


9
.
2

Data, Data, Data


A final important factor
is the explosion in data. In the past data volumes have been driven
primarily
by
larger models running on larger computers
; however,

the volume of data from other sources is now
beginning to dominate. Large experiments and instruments such as ATLAS and the

SKA a
re generating or
will generate petabytes to 100s of p
etabytes of data
. In addition,

many other activities are generating
data: gene sequencing, medical imaging, sensor networks, satellite data to name just a few. The sources
of digital data are growi
ng at a

faster

rate tha
n

is storage capacity
.

The provision of data storage and the
management and archiving of large datasets must become a key focus of Compute Canada. Some of
these issues are described in the following sections.


Data storage, managemen
t and archiving are continuing to grow in importance for HPC data
centre
s,
driven by the
current

"data explosion."

Data and
the
information
derived from it
are core

component
s

of science and as a result accessing, handling, sharing and combining that data
and information will
become more and more important.
D
ata
-
intensive applications are quickly emerging as a significant new
class of HPC workload
.


As a result of the digitization of the research enterprise a growing number of researchers must address
the
storage and processing of increasing amounts of digital data. This

data deluge


is increasing the
number of both traditional and non
-
traditional researchers that require access to large storage and
computing capacity of the type normally available
only
a
t HPC sites. They will need a
very significant

amount of storage, appropriate software
and computing

power to process and analyze the data sets
obtain
ed

from experimental platforms or clinical trials.

The International Review Panel suggested that Compute

Canada consider its role in the management of
research data and use the data storage capabilities and expertise of the
regional divisions

to develop a
national approach to managing large research data collections.

In order to support the full requirements

of researchers and scientists, Compute Canada should be
working with its partners to provide

a highly
integrated storage infrastructure that supports the
computational facilities
.





23

|
P a g e


Data Storage

Support will be provided for storage of large (hundreds of Te
rabytes to Petabytes) data collections in
network proximity to the compute resources needed to process these data
.


Data Management

Facilities and policies will be developed to support operational data management systems hosted at
Compute Canada sites.


Da
ta Archiving

Facilities and policies will be developed to support long
-
term (10 years and longer
, some data sets in
perpetuity
) archiving of large science and other datasets.


9.3

Programs and Outreach


C
ompute
C
anada

will

offer
training
programs that mak
e effective use of infrastructure, bring a full range
of disciplines into the use of HPC, encourage
researchers to “upgrade” to take advantage of
technological advances

and prepare for a
Tier 1

machine
, offer science fairs,
and work

with
SMEs and

the private s
ector generally.


SMEs are an important element within the Canadian economy, accounting for approximately 80% of
businesses.
Access to HPC resources and training will enable some SMEs to become and stay
competitive and to grow their global market share.
The International Review Panel report
recommended that Compute Canada develop a coordinated plan that would provide better support to
commercial organizations. SMEs may be aware of the competitive advantages of HPC
; however,

they
may lack the necessary ski
lls to take advantage of the technology and apply it to their businesses. Even if
they do have the skills, they may find it very difficult to access HPC equipment.

Encouraging Canada’s
small and medium sized enterprises to adopt and adapt
HPC

for competit
ive advantage and
sustainability will give them leverage in the marketplace.

Much research today requires HPC as a tool. The benefits of the research conducted and the resulting
achievements include dramatic reductions in the time to results ratio; underta
king research in new and
little
-
understood areas; the ability to access and analyze more data and more complex data sets; using
simulation rather than building physical models to construct airplane wings or crash
-
test cars. Compute
Canada will develop a st
rategic communications plan to ensure that Canada’s accomplishments in HPC
-
driven research and innovation are better articulated to all of our stakeholders. In addition, the major
science initiatives, as partners in research, will be asked to acknowledge t
heir use of and reliance upon
Compute Canada resources.









24

|
P a g e



10.0

Long
-
Range Budget Forecast

($millions)

Note: Each column represents a three year period

($Millions)


2009-2012
2012-2015
2015-2018
2018-2021
NPF
Projected
Projected
Projected
TOTALS
INFRASTRUCTURE
Mid-Level
72
180*
190
200
462
Tier 1
52
65
100
217
OPERATIONS
Staff
47
60
80
100
287
Software
12
18
25
50
105
Power
17
51**
60
75
152
Other (rent, etc.)
15
15
20
20
70
PROGRAMS
Outreach/non-
10
10
15
18
53
trad/SMEs,etc.
TOTAL
173
155
455
563
1,346


Notes:

*
This increase would bring Canada closer to
the mean of the G8 countries in terms of investment in
HPC.

* *
Assumes the regional divisions will pay the full cost of power when data centres are rationalized.


To bring the Canadian HPC/GDP ratio to the mean for the industrialized countries w
i
thin a
five year
period would

require

an investment in HPC
equipment

of $
72

million

per year or $360 million
.

An
investment of this order

would place Canada between the UK and Germany in terms of the HPC to GDP
ratio. (It is worth stressing that this would not me
an that we would have nearly as much HPC in absolute
terms as the UK since our GDP is less


in fact we would have around 40% as much


but the availability
per GDP$


or re
searcher


would be much closer
)
.

This level of investment would
place

Canada
betw
een Taiwan and Saudi Arabia
.



1
0
.1


Consequences of Underfunding


There will be significant

c
onsequences of
u
nderfunding

the national HPC platform and the support
Compute Canada provides the Canadian research community. These consequences will
directly im
pact
Canada’s

achievement
s

in science

and our international reputation
, economic performance and the
ability to innovate
. These include:


Brain drain to those countries that are supporting HPC (scientists, researchers and the highly
qualified personnel who

support them)
;


Researchers
will

limit

their research due to the limitations of facilities
;


25

|
P a g e



Difficulty in attracting
or retaining
Canada Research Chairs and Canada Excellence Research
Chairs
;


Competitive disadvantage in research and in industry and innovat
ion
;


Loss of ability to participate in the grand challenges (climate change/pandemics/energy/ATLAS)
that are international collaborations
;


Lost opportunities for researchers in disciplines such as the humanities and social sciences who
are beginning to dev
elop and leverage HPC to their advantage
;


Movement of researchers to private or commercial clouds even if these do not fully meet their
needs (i.e. our researchers settle for second or third best)
;


Decrease in R&D investment in academic research by the pri
vate sector
; and


The loss of scientists and researchers who rely on mid
-
level

capabilities and support.



1
1
.0

Performance Assessment Framework

The performance assessment framework consists of two components: The Logic Model and
the
Performance Measuremen
t Model
.

The logic model in figure 2 will allow the Board of Directors to
evaluate the performance of Compute Canada and its regional centres.

A performance assessment framework is outlined in figure 3.

11.1

O
bjectives


11.1.1

To establish and sustain an
integrated
national HPC platform

that takes advantage of new
technologies and software in order to provide the best possible resources for researchers by:



Designing, acquiring, maintaining and managing efficient and cost
-
effective HPC facilities



Broadeni
ng the base of partners for
whom Compute Canada
host
s equipment and offers

shar
ed

resources



Establishing

a Tier 1 facility in Canada, maintaining mid
-
range facilities and ensuring support for
researchers who need routine access to HPC for research advantag
e



Improving the support for data intensive computing and d
eveloping the capability for
large scale
and long
-
term
data
management,
storage and archiving


11.1.
2

To provide expert
support for researchers

using HPC in order to enhance their efficiency and
eff
ectiveness and
to
accelerat
e

the results of their research by:



Optimizing access to and utilization of HPC facilities by researchers
irrespective of location



Developing HQP to meet present needs and prepare for future requirements



Offering a variety of pro
grams that encourage and facilitate the use of HPC




26

|
P a g e



Figure
2
: Compute Canada
LOGIC MODEL


National Platform and Researcher Support

Outreach

Develop Highly Qualified
Personnel

Infrastructure Expansion
,
Upgrade
&
Management

Selection
,
Purchase
&
Installation of HPC Capacity
&
Equipment

Infrastructure
,
Software and
Middleware Support

Access to
&
Allocation of
Resources
&
User Support

Program
&
Project
Management

Efficient
&
Effective
HPC Platform

Upgraded HPC
Infrastructure

Improved capability
for use of HPC
Infrastructure

Improved capability
for dynamic storage
,
sharing
,
manipulation
&
analysis of data

Improved access to
HPC Infrastructure

Improved access to
HQP

Increased national
collaboration

Increased
international
collaboration

More extensive use of shared HPC
platform

Increased number of HQP who can
support HPC

Enhanced profile for HPC in Canada

Strengthened Research Capacity

Strengthened Innovation Capacity

Strengthened Digital Economy and
Social Well
-
being

Special Programs
&
Projects

Objectives
Programs
Activities
National
Platform
Outputs
Researcher
Support
Outputs
Outcomes
Ultimate
Impact


27

|
P a g e


Figure 3: Performance Measurement
Model

OUTPUT

PERFORMANCE INDICATOR

DATA SOURCE

COLLECTION FREQUENCY

DATE TO ACHIEVE





















National Platform



















Efficient & Effective HPC Platform


# researchers utilizing










Processor clock speed, peak
flop rate, cache structure,
cache latencies, main
memory bandwidth and
latency, cores per compute
node, communic
ation
bandwidth and latency and
I/O bandwidth










Compare CC cost to buying

In the commercial cloud







Upgraded HPC Infrastructure

1 machine in top 20 of
TOP500

TOP500 list

semi
-
annually





machines replaced at 5 years









meets internatio
nal standards

















Improved capability for use of

increase in non
-
trad users

CCDB

annual




collaboration




HPC Infrastructure

increase in medical research

CCDB

annual























Improved capability for dynamic
storage,









sharing, manipulation & analysis of
data





























Researcher Support



















Improved access to HPC infrastructure

100% of large projects receive

NRAC

annual





allocation requested









user satisfaction

survey

annual





Improved science quality








Improved access to HQP

user satisfaction

survey

annual





75% of gap areas staffed










Increased research
productivity








28

|
P a g e


Increased national collaboration

80% of "big
-
science" served









Papers

and citations










SSHRC/NSERC/CIHR/CANARIE

grants supported by HPC









# of MITACS placements





# of SMEs supported




Increased international collaboration

Papers and citations
















12.

Conclusion

The success of Canadian resear
ch depends upon
our researchers having access to the mandatory tools
for research in the twenty
-
first century
.
In an era of massive data volumes and the need to understand
complex phenomena,
High Performance Computing
has become

an indispensable tool for t
he
advancement of knowledge and the resulting innovation, wealth creation and well
-
being for Canadians
.

With the capabilities afforded by HPC resources, s
cientists and researchers
are
a
ddressing issues of
immediate significance for contemporary society.

T
hey position the Canadian government to make
informed policy decisions; they position the medical community to
keep Canadians healthy; they can
position the financial community to
understand and
avoid economic meltdowns such as we have
recently experienced
. They make our world and our lives richer and safer.

Compute Canada
will ensure that

our stakeholders, including all
Canadians
,

both understand and reap
the benefits

of the research that results from our provision of world
-
class
HPC
systems

and human
re
sources to Canadian scientists and researchers.

















































































29

|
P a g e


APPENDIX
1 : International Review Panel Corrective Measures and Key
Recommendations

CORRECTIVE MEASURES


Based on its findings, the IRP r
ecommended to CFI that the following corrective measures should be undertaken
and implemented by Compute Canada to build on its current achievements and ensure the long
-
term success of
the Compute Canada initiative. A strategic plan for Compute Canada that

strengthens its national role in the
Canadian research landscape. The plan should contain:

o

The value propositions for organizations to participate in and contribute to Compute Canada.

o

A performance assessment framework and arrangements with the consortia

that would allow
the Board of Compute Canada to regularly review the performance of Compute Canada in
delivering on its plan.


A formal executive process to allow the Compute Canada Board to implement the strategic plan at both
the national and regional l
evel.

The IRP recommended that Compute Canada be asked to provide this plan to CFI by December 31, 2010.

Future CFI payments associated with project #12866 should be contingent upon the implementation of these
corrective measures in a timely fashion.

KEY

PANEL RECOMMENDATIONS


The key recommendation of the IRP is reflected in the corrective measures outlined above.

In addition, the IRP believes the following key activities should be undertaken by Compute Canada in the short
term:


An overall lifetime cost
-
benefit analysis of the acquired systems to inform future infrastructure
acquisitions.

o

Compute Canada should conduct a detailed cost
-
benefit analysis to determine which sites should
eventually receive additional infrastructure. This analysis should take i
nto consideration the cost
over the lifetime of the equipment (acquisition + on
-
going operation). It should also consider any
other benefits of placing infrastructure in a given location (redundancy, contribution from an
institution, etc.).


A gap analysis
for the need of additional support personnel

o

While there is general agreement that additional support personal would be beneficial to users,
there is a need for a gap analysis of the skills required to support the user community. This
analysis should clea
rly outline the costs and benefits of having more support staff.


Regarding the NRAC, Compute Canada should:

o

Explore arrangements to provide long
-
term allocations and support to the most meritorious
research groups.

o

Raise the awareness of the process and i
ncrease the number of applications by undertaking
promotional activities.


A coordinated plan that would allow the consortium members to better support non
-
traditional
communities and commercial organizations.



30

|
P a g e


1.

Cost
-
Benefit Analysis

All
regional divis
ions are contributing to the cost of this analysis which is currently being drafted. The
University of Alberta has agreed to act as the contractor on behalf of Compute Canada. Compute
Canada will review all responses to the request for proposals which will

be issued in early January. CPAC
has developed the following principles for the location of equipment and personnel which will be used
to guide the analysis:



1.
Accountability:
Centres will operate in close, ongoing coordination with CC and will adhere

to agreed national standards and practices. Past performance will be crucial to determine where


new resources will be located.



2.
Objectivity:
Compute Canada should set up an external review process to insure that the


objectives are met in a rational

way.



As a consequence:



‐Individual data centres should be reviewed by an external review panel.



‐The review should be based on past performance and on a sound proposal for future



hosting. An example “application form” for individual data centres is included as an



annex.


The full text of the document may be found on the Compute Canada website
:
(
www.computecanada.org
)
. It is anticipated that the work for this will take approximately 60 working
days from the date the contract is

issued and will involve site visits.


2.

Gap Analysis

This analysis is being undertaken internally and is intended to identify those areas in which new and/or
additional user support skills and the corresponding staff are required to meet and anticipate t
he needs
of researchers. Areas which will be reviewed include:



data management and archiving;


the specialized skills required to support medical research within a shared national platform;


support to the arts, humanities and social sciences in order to m
igrate those researchers from
the research computing level to the true high performance computing that will advance their
research and ensure they are internationally competitive; and


the development of the skills, both maintenance and software, required i
n preparation for
exascale computing and Compute Canada’s ability to assist researchers to move to exascale
computing quickly.


3.

NRAC


The key recommendation as stated by the International Review Panel has been fully met.


As stated in the September 2010

Call for Proposals “The primary mission of the NRAC is to ensure that
the best science that is received in the proposals is put on the appropriate machines.” The call indicates
that allocations will be valid for one year; however, it also notes that “Unde
r exceptional circumstances
and for
long term national and international projects an extended timeframe may be granted.

31

|
P a g e


Allocations of longer than one year will be subject to NRAC approval and will require the submission of
an annual progress report.”


In
order to ensure awareness of the process, Compute Canada promoted the fact that the call would be
issued in advance of the release date. Notification was sent to everyone holding a Compute Canada
Database (CCDB) account. In addition, the regional divisions

distributed emails to a wider audience and
included prospective applicants even if they are not currently in the CCDB.


4.

Coordinated plan to support non
-
traditional communities and commercial organizations.


There are three elements to Compute Canada’s

plan to support commercial organizations:


i.

A Memorandum of Understanding has been signed with MITACS to identify a minimum of 10
internship opportunities. This will provide not only support to commercial organizations but will
be designed to advance the p
otential for using HPC in their businesses and encouraging further
investment by the private sector in R&D.

ii.

Providing access by SMEs to a limited amount of Compute Canada high performance computing
equipment and expertise.

iii.

A Memorandum of Understanding wi
th Rocky Mountain Supercomputing Centres in Montana to
develop virtual clusters of expertise (wind power, precision agriculture, manufacturing) for SME’s
in order to develop new products.


There are four primary elements to Compute Canada's plan to support

non
-

traditional communities.


i.


Development of a comprehensive support resource for identified communities.


This would include: web resources that were widely advertized to target communities that

detailed clearly the range of support available; ded
icated support personnel, knowledgeable in

these disciplines
-

ideally some of these staff would be new hires; and a clear path indicating

how these resources, including the specific additional elements of the plan described below, can

be accessed.

ii.


Provision of alternate computing models, including support of portal interfaces to HPC and

cloud computing, that are often more closely aligned with the needs and research

methodologies of the non
-
traditional user communities.

iii.


Targeted conference
s and workshops. Conferences and seminars expose the community to

examples of what has been achieved in these disciplines by providing successful examples and

allow for exchange of ideas. Hands
-
on workshops, fully supported by technical staff, have been

successful in moving projects from the ideas stage to a prototype within a few
-
day intensive

session.

iv.

Targeting a small number of key projects
-

selected through a competitive process
-

for

accelerated support. This may typically involve programmin
g support to develop applications for

HPC platforms, data visualization etc.


All of these strategies have been used within various areas of Compute Canada already. This plan
formalizes many of the successful activities.
It is anticipated that

the implem
entation
of this plan
will be
that some significant aspects are undertaken as a specialization within one or two of the regional
divisions.


32

|
P a g e


Appendix
2
:
What
Researcher Stakeholders
Are S
aying

1.

CANADIAN ASTRONOMICAL COMPUTING, DATA AND NETWORK FACILIT
IES: A WHITE PAPER

FOR THE 2010 LONG RANGE PLAN


L. Jonathan Dursi (CITA), David A. Bohlender (HIA/NRC), James Wadsley (McMaster University)

and JJ

Kavelaars (HIA/NRC)


ABSTRACT

Significant investment in new large, expensive astronomical observing facili
ties spanning a

sub
stantial portion of the electronic spectrum was a dominant theme of LRP2000 and continues to be

necessary for Canadian astronomy to maintain its world position. These developments are generating

increasingly large volumes of data. Such
investments only make sense if they are balanced by strong

infrastructure support to ensure that data acquired with these facilities can be readily accessed and

analyzed by observers, and that theoreticians have the tools available to simulate and understa
nd

their context. This will require continuing investment in computational facilities to store and analyze

the data, networks to ensure useful access to the data and products by Canadian researchers, and

personnel to help Canadian researchers make use of t
hese tools.


In addition, large parallel simulations have become an essential tool for astrophysical theory, and

Canadian Astronomy has world
-
leading simulators and developers who rely on world
-
class High

Performance Computing facilities being maintained
in Canada to do their research e
ff
ectively.

We recommend that Compute Canada be funded at $72M/yr to bring HPC funding per capita in

line with G8 norms; that part of every Compute Canada technology renewal include a Top
-
20 class

computing facility; NSERC a
nd other funding agencies begin supporting software development as an

integral component of scienti
fi
c research; that the sta
ble

funding for consortia be tripled, including

local access to technical analyst sta
ff
; and that the last mile bottleneck of campu
s networking less

than 10 Gb/s be addressed where it is impacting researchers, with particular urgency for the current

1 Gb/s connection at the CADC.


2.

CBRAIN

CBRAIN is a national
HPC
network for brain imaging research into brain disorders of aging (e.g.
Alzheimer’s Disease), development (e.g.
a
utism) or social ills (addiction, stress). Researchers use brain
scanners (MRI or PET) to capture 3D or 4D
phenomic data that is then subjected to extensive
computational analysis
at each 3D location
. This involves spatio
-
temporal models of physiological or
anatomical

c
hange

that present

similar mathematical problems to those of
ma
ny other HPC
-
focussed
domains
such as

astronomy or meteorology


except
that the analysis

has to be analyzed in 1000s of
individual
brain datasets and statistical models developed to characterize the phenomic differences
between normal brain and disease. Our challenge is to conduct e
ver more complex analysis on ever
more detailed 3D brain maps.


For instance, we are now operating with datasets that are 10K samples in
each dimension, i.e. 1TB per dataset. Increasingly, brain researchers are seeking to integrate these
phenomic maps wit
h genomic information to explore the gene
-
environment origins of bra
in disease in
transgenic or knock
-
out mouse models, where one can generate databases in the 100 TB range. Two
examples of current challenges illustrate the HPC challenge, (i) connectivi
ty analysis which explores the
covariance of

any two 3D locations in the brain

(an N
2
problem
)
, or (ii) genome
-
wide assiocation

33

|
P a g e


(GWAS) studies at each 3D location. Conducting such analysis

on 10K
3

datasets

is beyond curren
t
Compute Canada resources. I
t i
s essential that Canada’s HPC infrastructure keep pace with the
international field in these areas. CBRAIN is now connected to similar international efforts in Europe, the
US, China, Korea and India as we build a global network of HPC partners engaged in

brain mapping,
the
GBRAIN

initiative
. Canada

is well
-
positioned to be a global leader within this international brain
mapping network and to develop innovative strategies that combat brain disease
-

as long as Compute
Canada continues to provide world
-
cl
ass HPC support.

3
.

Canada’s Digital Environment for Research, Innovation and Education, a submission under the

Digital E
conomy Strategy Consultation by Canadian Digital Media Network, Canadian Research

Knowledge Network, Canadian University Council of CIOs, CANARIE Inc. and Compute Canada

The elements of the

digital
environment to support
RIE
[Research, Industry and Educati
on]

have b
een
evolving and

include:
the preservation and management of
huge repositories of data and rich digital
content; ever
-
larger compute capacity; digital devices and distributed sensors; low
-
laten
cy, high
-
bandwidth networks;
middleware that integrat
e
s

the infra
structure and supports its use; and the
expertise required to manage and operate them. …
The elements of Canada’s
digital infrastructure

for
RIE have been evolving for several years, enabled by increasingly powerful computing and networking
tec
hnology. Canada’s investments in these areas are part of a world
-
wide response to the growing
reliance on ever
-
increasing volumes of
shared

research

data
. A parallel shift towards greater

reliance on
collaborative models

is
a response to the flood of da
ta and the collective need to
manage it
in a cost
-
effective way.
The combination of highly skilled personnel, collaborative models, and appropriate
supporting infrastructure were identified as key elements supporting Canada’s innovation system in the
Sci
ence Technology, and Innovation Council’s 2008 State of the Nation Report.

4
.

Resolution concerning High
-
Performance Computing and the Humanities

Be it resolved that:

The Society for Digital Humanities / Société pour l'étude des médias interactifs (SDH/SEM
I) recommends
that the Society work with Compute/Calcul Canada (CCC) to develop the high
-
performance computing
(HPC) facilities funded for Canadian researchers so that they can be meaningfully used by humanists in
need of research computing support. To ach
ieve meaningful engagement SDH/SEMI encourages CCC to
include digital humanists in their planning and governance process in order that the next generation of
CCC facilities, support, and training be appropriately extended to be truly inclusive. SDH/SEMI of
fers to
work closely with CCC to make the case for support of humanities research.

Further SDH/SEMI recommends that universities and funding bodies like SSHRC respond to CCC
engagement with programs capable of supporting research using CCC facilities adequ
ately where that
research meets standards of excellence.

Background
. In the last two decades there has been an epochal shift in humanities research as the
cultural record we study and care for is being digitized. The mass
digitization of human histories,
l
iteratures, art, cinema and music has

challenged our research practices deeply but also offer the
opportunity for large
-
scale digital humanities. Compute/Calcul Canada could be a partner of the
humanities as it changes the scale of its work and learns to a
rticulate new types of questions. Also, given
that the humanities represent a significant percentage of the professoriate, the serious engagement of

34

|
P a g e


CCC in supporting research across the arts and humanities would dramatically increase the pool of users
of
the CCC infrastructure. Further, the materials humanists work on interest all Canadians


after all, we
study the arts, the stories, and the histories that Canadians care passionately about and use to define
themselves as Canadians. Support for humanities
research would therefore have secondary effects that
benefit all Canadians. Imagine Compute/Calcul Canada playing a foundational role in Canada’s cultural
knowledge; SDH/SEMI can!

Two HPC Consortia, SHARCnet and WestGrid, have reached out to the digital hu
manities community
organizing special events and acting on recommendations. See the draft report from the recent Mind
the Gap workshop for background and links,

http://docs.google.com/View?id=d
hbw7427_4hnbkr8cd

.
SDH/SEMI encourages CCC to take the recommendations coming out of such meetings seriously. We
realize that many of the recommendations propose new services that might dilute the core mission of
CCC, but we believe that that core missio
n can be extended and that extending CCC is preferable to
developing parallel infrastructure. We believe CFI thinks so too. For all these reasons we encourage CCC
to engage the humanities in a meaningful way and offer to assist.

5
.

ATLAS

15 October
2010

CERN Computing Resources Review Board Summary

Walter Davidson (NRC), Canadian C
-
RRB representative

Robert McPherson (UVic/IPP), Canadian National Contact Physicist to CERN/ATLAS

Worldwide LHC Computing Grid (WLCG) operation with first data:
The initi
al use of the W
-
LCG has
been extremely successful, exemplified by the number of physics results extracted by the LHC
experiments in timely fashions from the earliest data. Included in the successes is the routine analysis of
data at the Tier
-
2 centres by
thousands of physicists using computing grid tools. Currently, over 100M
computing jobs are completed in the WLCG each day, corresponding to 100,000 CPU
-
days used each
day. Sites are learning from experience with first data which should lead to some econ
omies in the
future. The load on operations personnel at the centres is high, but just manageable by making strong
use of collaborative problem solving using experts from different sites. It is clear that we cannot reduce
the number of people running the

centres. Detailed scrutiny of the computing use by the different
experiments show that storage allocations are used efficiently, while some inefficiencies are present in
the CPU use at the Tier
-
0 and Tier
-
1 centres. Gratifyingly, ATLAS stands out has hav
ing Tier
-
0 and Tier
-
1
CPU use efficiency nearly a factor of two better than the CMS experiment. CMS funding agencies
particularly encouraged economies be found by increasing the utilization efficiency.

Resource planning for near and medium
-
term future:
cou
ntries and funding agencies are committed to
future WLCG support. Eg, in 2011, ATLAS in particular has full commitments from all funding agencies
for our needed computing resources. All funding agencies encourage continued resource scrutiny
seeking econo
mies, including understanding the implications of the LHC running schedule which will
likely result in resource needs peaks and valleys. While the schedule may result in modified purchase
profiles, the total running time averaged over a few year period wi
ll be the same as that used in
computing resource need projections so the total resources needed will not change significantly. All
funding agencies are committed to maintaining their share of computing resources required to analyze
the LHC data.


35

|
P a g e


Summar
y for Canada:

we should expect a continued need for computing resources for ATLAS in Canada.
There will be some year
-
by
-
year fluctuation due to the LHC running schedule with planned shutdowns in
2012, 2016 and 2020 but with continuous operation (multi
-
yea
r runs) through other years. Our
international partners are maintaining their strong commitment to the WLCG, and maintaining our 5%
share of ATLAS computing will require computing hardware resource funding at the level of our existing
projections, as well

as maintaining existing support for computing centre and grid operations at the
Canadian sites. For the TRIUMF Tier
-
1 centre, solutions for both hardware renewal and centre
operations after the CFI award completes in early 2012 are critical to continued
Canadian exploitation of
our LHC investment
. For the university
-
based Tier
-
2 sites, continued allocation of computing
hardware resources from Compute Canada and operational support from the universities, Compute
Canada and NSERC funds is essential
(emphas
is added)
.

6
.

O
ntario
C
ancer
B
iomarker
N
etwork (OCBN)

The close collaboration that the OCBN has established with Compute Canada enables OCBN to function
seamlessly as a single entity across the multiple nodes of the organization, many of which are conne
cted
in a virtual manner. With the high performance computing support and infrastructure provided, the

OCBN is able to execute in parallel multiple, large‐scale research programs, each of which has significant

computational, storage, and workflow managemen
t needs. Without the infrastructure and support of

Compute Canada, the OCBN would be severely limited in its ability to carry out its business and scientific
activities to the level required to be competitive nationally and internationally. Further, OCBN w
ould be
severely compromised in its ability to coordinate and support the activities of its multiple proteomics
and genomics core laboratories without the expanded collaboration between Compute Canada and
OCBN.


7
.

SNO
LAB


Researchers at SNOLAB institution
s, including Queens and
other
institutions

in Canada and
internationally have been using
Compute Canada resources

for the detailed Monte Carlo simulation of

detectors under development as

well as extensive analysis of detectors in operation. The scientific

objective of these detectors is the observation of dark matter particles created in the Big Bang that are
known to make up about 25% of the Universe. Another objective is the observation of neutrino
-
less
double beta decay that can provide information abou
t neutrino mass and processes in the early
universe that led to our matter
-
dominated universe. Other measurements will include neutrinos from
the sun, the earth and distant supernovae. All of these measurements will benefit from the unique
environment at S
NOLAB that is the lowest radioactive location in the world. HPC is essential for all of
this work as the complexity of the analysis and simulation work requires such capability.
” (Dr. A.
MacDonald, SNP Project Director)


8
.

NEPTUNE Canada and VENUS

"The NE
PTUNE Canada and VENUS cabled ocean observatories are world
-
leading facilities in a new
generation of ocean S&T which brings electrical power and the Internet to the ocean enabling
continuous and concurrent measurement of a broad suite of biological, physi
cal, chemical and geologic
properties in ways hitherto not possible. The data streams, including scalar, video and acoustic data,
generate short and longer term time
-
series which support the analysis of complex interactive ocean
system processes. But herei
n lies a major high performance computing challenge, namely the

36

|
P a g e


management, integration and analysis of massive data sets of varied modalities over time. These
challenges find expression in areas that include, but are not limited to, data visualization, ma
ssive data
handling, adaptive modelling, systems modelling, and non
-
parametric analysis. At the same time these
challenges are exciting opportunities for harnessing leading edge ocean science with the HPC
capabilities provided through Compute Canada, all s
upported by the CANARIE network. As a result,
Canada is especially well positioned to be an international leader in this area."

(
Dr.
M. Taylor,

President
& CEO, Ocean Networks Canada)



37

|
P a g e


APPENDIX
3
:
Stakeholder
Comm
e
nts on the Digital Economy
Strategy
Consu
ltation

Submission


This is an excellent suggestion. HPC does change the way we think about research.

All of a sudden we can do myriads of things at the same time unraveling

fundamental principles hidden in large arrays of data. Access to HPC is critical

t
o so many areas of our life spanning from pure science such as astrophysics and

particle physics to weather modeling, genome/drug research to day
-
to
-
day banking.”

(noskovsy)

“I am an academic psychologist who studies how human memory works. The work has

bo
th theoretical and practical implications. Applications include knowledge

retrieval & translation. Without HPC, the field is dead in Canada, and my

students will continue to contribute to the US economy. With HPC, my students

can stay in Canada.”

(mewhort)

“HPC is a primary concern in my research. I carry out Finite Difference Time

Domain (FDTD) simulations of electromagnetic wave

propagation/scattering/radiation with an

in
-
house developed software with unique capabilities. The range of applications

is wide
spread, from design/analysis of antennas, photonics components,

electromagnetic interference, metamaterial, etc. HPC has opened new possibilities

in my research. I would not think of doing without it now.”

(Jasmin E Roy)

“HPC as long been realized through
many international initiatives as a means of

increasing competitiveness. Canada needs to embrace and promote what has been

developed through Compute Canada. We have to provide appropriate access to this

Canadian infrastructure asset (cyberinfrastructure) t
o the research, innovation

and educational community. That means access even to the SME innovation engine of

Canada which makes up the highest collective economic driver we have.”

(TerryDalton)

“HPC is a key to success in any scientific research these days
.

Unfortunately, Canada is way behind all other developed countries in this

competition. I vote that it should be changed.”

(sergeychelsky)

“Without HPC most of the problems we solve are of the nature of conceptually

idealized scenarios only. HPC makes mas
sive simulation possible, which allow us

the capacity to deal with real field problems.”

(zhaoga)



In genomics/proteomics/bioinformatics research HPC pervades at multiple levels
-

from storing and crunching large volumes of raw data generated from next
-
ge
n

sequencers and mass spectrometers, to enabling data integration and data mining

of these datasets, to providing security and privacy with respect to confidential

clinical data. This infrastructure is critical to retaining our highly qualified

talent righ
t here in Canada, and in attracting international business and

investment in our biotech and research sectors. The need for HPC


the hardware,

the software, and the people


will only continue to grow, particularly in light

of the “grand challenge” of Per
sonalized Medicine, which is as much a


38

|
P a g e


computational problem than anything else, given the complexity of the molecular

mechanisms of biological systems and diseases. We have an opportunity to be at

the forefront of this research, and investment in HPC will

be critical to getting

there.

(mdharsee)


“One aspect of computing that is changing the economics of product development is

the use of extensive simulation instead of prototyping ( The Boeing 787 needed

only 4 test wings instead of 18 for the 747). Simul
ation is becoming the key

factor in getting a competitive advantage.

Extensive simulation requires large scale

computing: that means large scale

computing infrastructure BUT ALSO and more

importantly top level programming

specialists. One cannot do witho
ut BOTH.

Because of these essential
requirements, developing

countries cannot compete with

their cheap manpower on that innovation front, hence a clear advantage for

countries developing a computing literate workforce ( here I am not referring to

window us
ers but advance programming specialists). This is one area where Canada

can best exploit its university trained students.”

(jmpoutissou)


“Well supported HPC infrastructure and associated human support is important also

if Canada is to develop its digital
content industry (from animation to games to

electronic texts.) HPC facilities will be the backbone to large
-
scale archives,

large
-
scale content mining, and multimedia content sharing.”

(GeoffreyRockwell)

“Increased HPC investment, to support SME sector, c
an provide an immediate

advantage to Canada. If we look at one sector, renewable energy, and more

specifically wind, an SME with access to an HPC to assist in the selection of a

wind farm site, can make a site selection that generates 25%
-
50% more power th
an

a site selected without access to HPC’s. The result is a competitive SME, HQP

job growth, and a greener economy.”

(kmacneill)


39

|
P a g e




APPENDIX

4
: HPC/GDP


HPC/GDP (2005-2009 average)
0.03
0.05
0.06
0.07
0.14
0.17
0.18
0.19
0.20
0.20
0.21
0.21
0.22
0.25
0.28
0.28
0.32
0.37
0.38
0.39
0.53
0.54
0.54
0.56
0.56
0.59
0.61
0.63
0.64
0.65
0.67
0.69
0.72
0.76
0.79
1.24
1.26
1.27
1.50
2.34
2.46
0.00
0.50
1.00
1.50
2.00
2.50
Turkey
Indonesia
UAE
Mexico
Brazil
South Africa
Egypt
Austria
Belarus
Belgium
Denmark
Singapore
Ireland
Luxembourg
Italy
Poland
Australia
Canada
Korea (S)
Russia
Cyprus
Netherlands
China
Japan
Spain
Norway
Saudi Arabia
Taiwan
Malaysia
India
France
Finland
Slovenia
Bulgaria
Germany
UK
Sweden
Switzerland
Israel
NewZealand
US
HPC/GDP (relative to mean of "G6")
G8 U G10 (excluding Canada)
Canada
Other advanced/developed nations
from comparator group (Fig 1)
Other HPC-invested nations
Fiducial factor of 2 above and below mean of "G6"


40

|
P a g e


APPENDIX
5
:
Strengths, Weaknesses, Opportunities and Threats


Strengths



Pan
-
Canadian national HPC pl
atform being consolidated and integrated with substantial
infrastructure and researcher access to machines and expertise across Canada



HQP, solid base of HPC expertise



Distribution of user support staff to allow close collaboration with research teams



Supp
ort for “big science” projects/international collaboration



Enhances

federal and provincial investments in research



Enabler for innovation



On
-
going examination of technological advances in equipment and software development to
ensure researchers needs can b
e met



Excellent university research using HPC



Acquisitions are cost
-
effective and represent good value for the funds available (IAP Report)



Contribution of committees/regional staff to Compute Canada



Ability to coordinate activities at the national level



Q
uality facilities and support provided within a very limited budget


Weaknesses



The governance structure needs to be strengthened



Insufficient staff

and resources

in Compute Canada
’s central office



Insufficient user support personal



Lack of a Tier 1 facili
ty



Limited outreach to non
-
traditional research disciplines



Few connections between Compute Canada and industry



Low profile of Compute Canada



Researchers/scientists have been diverted to the management of HPC facilities



Dependence on “volunteers” to serve

on committees reduces the time scientists have for
science



Lack of service standards for data centres



Lack of a robust business model


Opportunities



Industrial outreach through training courses, workshops and collaboration with academic
researchers



Devel
opment of a national approach to managing large research data collections in conjunction
with other national organizations



Increased use of HPC in the humanities and social sciences



Creation of a shared environment that meets the needs of the medical resea
rch community



Convince CFI to aggregate small HPC grants to expand and sustain the national platform



Green HPC/energy efficiency



Increase collaboration/partnership between research and industry, particularly support to SMEs
to demonstrate economic impact



I
mplementation of a Tier 1 facility and readiness for the arrival of exascale computing;



Create career progression opportunities for HQP and build Canadian expertise



Implement a cloud computing test bed


41

|
P a g e


Threats



Increasing cost of power



Limited political s
upport for or recognition of the strategic importance of HPC to research and
the economy



Funding is significantly lower than that of our trading partners
and Canada is falling behind in
providing HPC to support research



Dependency upon multiple sources of
funding creates several different sets of rules and
accountability, making integration difficult



Lack of sustained, long
-
term funding



Lack of funding for HPC research and limited resources to research HPC systems, hardware,
software, middleware and trends



Loss of HQP/ brain drain to other countries where HPC is a priority and well
-
funded



Funding for HPC infrastructure and personnel unknown after March 31, 2012



Lack of integrated federal and provincial funding for HPC



HPC equipment is very expensive

and has
a short useful life. Almost all information technology
equipment grows obsolete before it stops working. The well
-
known Moore’s law states that
computer performance, or “bang for the buck”, doubles every two years



Movement of researchers to private or com
mercial clouds if Compute Canada funding is
insufficient to provide the resources they require



42

|
P a g e



A
PPENDIX 6
:

S&T Priorities
: Examples of
Support by
Compute Canada


Research Area: Environmental Science and Technologies

Research Objective

The
reduction of
carbon emissions
given its

economic and societal
priority

and significance
in Canada

through better

understanding and

o
ptimiz
ation of

fluidized bed catalytic gasification. Gasification

refers to the conversion of solid or liquid carbonaceous material to a

clean synthesis gas (syngas) product. Gasification can produce

hydrogen, which can be used for green energy sources such as fuel

cells, while also producing a concentrated carbon dioxide stream

suitable for capture
-
and
-
storage sequestration. In a world wi
th high

fossil fuel prices and low emissions targets, it is expected that

environmentally friendly gasification of low
-
grade coals and other

biomass or organic waste will become an increasingly important

technology in the production of green energy sources
.




Objective with current resources

Despite the astounding capabilities of modern computing, producing

high
-
fidelity simulations of industrial
-
scale gasifiers remains

impractical without specialized computing equipment.
E
ven using
relatively new equipme
nt from Compute Canada and the

state
-
of
-
the
-
art US Department of Energy software MFIX, it typically

takes 24 hours of real (clock) time to produce simulation of 50

seconds on a model domain that is ten thousand times smaller than
a real industrial gasifier
. Furthermore the anticipated memory
required for a realistic simulation is much larger than what is
available
.

Performance numbers such as these make it impossible to
take full advantage of the power of simulation for practical purposes.


Objective with
Tier 1

Petascale computing will allow
the production of

realistic data that
can be compared to experiment to further
improve mathematical

models. It

will also provide a framework for
the development

of

better simulation algorithms and ultimately perform pr
ocess
optimization
.


Potential Application

T
he transfer of knowledge to industrial partners.











43

|
P a g e


Research Area: Environmental Science and Technologies

Research Objective

The problem of global climate change and climate change projection
is a widely
acknowl
edged "grand challenge problem"
. Projections of
the ongoing impacts of the global warming process into the future
for a century or more are required input to the formulation of the
appropriate national and international policies that are needed to
m
inimize harmful impacts on the planet’s life support systems.

Objective with current resources

Present capabilities of computer systems in the fraction of a petaflop

(1000 trillion calculations per second)

range enable global
projections based upon assume
d
R
epresentative Concentration
Pathways (RCP's) for greenhouse gas increase, the increase to be
performed at a horizontal spatial resolution of approximately 100
-
200 km. This resolution is far to coarse to serve as a basis for
regional environmental policy

formulation. At present
computing
systems

are able to produce useful finer scale projections only
through the application of dynamical downscaling techniques. A
further issue in climate change projection is the fact that existing
models currently do not i
nclude a number of system components
that are expected to be crucial on the century timescale over which
useful projections are required. These include sub
-
surface hydrology
required to capture the impacts of climate change on the depth of
the water table
and the additional climate response associated with
the reaction of the great polar ice
-
sheets to the warming process.
Both of these hydrological impacts are seriously compromised by
inadequate spatial resolution of the model
s.

Objective with Tier 1

The a
vailability of true Tier 1 capability will enable global scale
models to be integrated at the high spatial resolution required to
address these critical issues without the need to invoke dynamical
downscaling techniques which are significantly error prone.

Potential Application

National and international policy makers will have reliable
information with which to make the decisions required to address
issues related to climate change impacts and adaptation.















44

|
P a g e


Research Area: Natural Resources and En
ergy

Research Objective

Associated with meso and microscale modeling of wind flows is the

application to wind turbine aerodynamic modeling (so power
performance

of individual turbines and windfarms
-

especially
interactions of

turbines with the wakes of oth
er turbines in a
particular site) and

noise generation.


Noise (both broadband and
tonal) is one of the

biggest hurdles for widespread application of
large wind turbines.

Being able to model the noise generation can
lead to design changes

that reduce turbi
ne noise.


Being able to
model and understand the

propagation of the noise from the turbine
to the surrounding ground

can lead to better regulation, limits and
zoning.

Objective with current resources

With current resources we can model

turbine aerodynami
cs and
power

performance, because this needs only a fairly coarse
resolution (still

on the order of 100 million control volumes, and
weeks with

hundreds

of processors).

Multiple runs to evaluate
changes and optimization

are unfeasible.


Modeling windfarms

and
wake interactions in real

settings is possible only by crudely
representing a turbine as a

simple disk that extracts power from the
local wind
-

all effects such

as rotation in the wakes of the turbines in
a windfarm are neglected.

We are stretched

to

be

able to

do the
very smallest computational

aero
-
acoustics simulations of the
simplest geometries

(weeks with

thousands of processors for a
simple flow impinging on a flat plate).

Computational aero
-
acoustics
of turbine blades is beyond current

resource
s.

Objective with Tier 1

With Tier 1 resource, we can do optimization of wind turbine blades

for power production, and interacting turbines in a windfarm
w
ithout

gross simplifications
.

We

would be able to do computational

aero
-
acoustics of turbine blades
and gain understanding of noise

generation mechanisms and

how the noise propagates.

Potential Application

Wind farms would be placed to maximize power output while
ensuring the health and safety of people living near them.














45

|
P a g e


Research Area:
Natural R
esources and Energy

Research Objective

Turbulence is a multi
-
spatial and time scale phenomenon that
pervades industrial and environmental air and water flows. It alone
remains as the last unsolved problem in classical physics. In
combination with combustio
n, the necessary predictive capability is
severely limited because current models can only predict NOx, CO,
and soot emission
trends

not levels, and cannot accurately predict
turbulent burning rates as turbulence intensity increases, and are
not able to pr
edict combustion instabilities for realistic,
technologically relevant configurations.

Objective with current resources

Existing HPC facilities do not currently permit high
-
fidelity simulation
of the entire complex geometry of full
-
scale practical combust
or
systems under realistic operating conditions (pressures,
temperatures, and turbulence levels), including coupling with other
engine and/or burner components without significant use of in many
cases greatly simplified modeling. For this reason, quantita
tive
predictions of unsteady, thermo
-
acoustic phenomena and
combustion instabilities are currently not possible. Such capabilities
are required for developing low
-
temperature, low
-
emissions burner
configurations. Additionally, today's high
-
fidelity simul
ations
can
not
provide quantitative predictions of most emissions and are either
too costly computationally or still not feasible to be used in formal
design optimization procedures for new combustor designs.


Objective with Tier 1

Access to a Tier 1 facil
ity would leverage existing Canadian research
expertise and accelerate the advancement of numerical combustion
science and enable Canada to be among the world leaders in the
development of the next
-
generation, high
-
fuel
-
efficient, low
-
emissions combustion
technology for conventional and emerging
alternative fuels. In particular, Tier 1 facilities would permit the
following:


DNS of larger laboratory
-
scale flames at higher more
practical levels of turbulence with more sophisticated and
accurate physical mode
ling, thereby performing simulations
containing more of the relevant physics, results from which
will yield a greater understanding and feed the development
of significantly improved and more accurate LES and hybrid
RANS/LES models.


LES and hybrid RANS/LE
S based simulations of full
-
scale,
industrial, combustor configurations at realistic turbulence
levels using more complete and more accurate sub
-
scale
physics models. Use of finer computational mesh (billions of
nodes) to alleviate the high reliance on mo
delling of
unresolved scales and provide quantitatively accurate
predictions of emissions such as CO and soot.


Numerical simulations of full
-
combustor configurations

46

|
P a g e


would be able to incorporate a full description of the
complex three
-
dimensional combustor
, including coupling
with other engine and/or burner components, and thereby
enable realistic and accurate predictions of thermo
-
acoustic
phenomena and the onset combustion instabilities.


The use of high
-
fidelity simulations in formal design
optimization p
rocedures for new combustor designs.


Potential Commercial
Application

In terms of fuel utilization, Canada has also become the international
leader in design and manufacture of small gas turbine engines for
aviation applications (Pratt & Whitney Canada)
and has made
significant inroads in the development of small and large gas turbine
engines for industrial and power generation applications (Rolls
-
Royce Canada). Pratt & Whitney's operations and service network
span the globe and its engines power the lar
gest fleet of business
and regional aircraft and helicopters. They employ 10,000 people
worldwide including 7,000 in Canada. In 1997, Rolls
-
Royce Canada
was designated the worldwide centre for design and manufacturing
of all large Rolls
-
Royce aero
-
deriva
tive industrial gas turbine engines
for power generation and they employ a highly educated workforce
of more than 1,000 people in Canada. To remain competitive,
Canadian gas turbine engine manufacturers will need to develop
new low emissions technology fo
r future advanced engines. They
will also require new combustion technology for alternative (non
-
fossil) fuels that will enable them to competitively enter new
markets that are expected to arise from the introduction of more
stringent environmental regula
tions. As an example of changing
policy, the International Air Transport Association (IATA) recently
outlined the aviation sector's commitment to environmental
responsibility, which includes the use of 10 percent alternative fuels
by 2017.
















47

|
P a g e


Research

Area: Health and Related Life Sciences and Technologies

Research Objective

The area of multiscale modeling of biological systems and processes
is advancing rapidly. One can envisage spatial scales going from the
picometer scale of electronic and atomic mo
tion in biomolecules
(nanometers) to the collective behavior of those molecules in a cell
(micrometers) and beyond to the formation of tissues (millimeters)
from cells, organs (centimeters) from tissues, organisms (up to
meters) from organs and ecosystems
(kilometers) that involve
organisms interacting with each other and with an environment.
Similarly large ranges of time scales apply, from femtosecond
chemistry, to nanosecond molecular dynamics, to millisecond
conformational changes of biomolecules, to t
he beating of a heart,
and so on. Researchers are currently examining small portions of
each of the space and time scales. The next great challenge is to
combine these approaches into integrated multiscale models of
biological phenomena.

Objective with cu
rrent resources

Let us take, as an example, the lower end of the spatial scale that
runs from atoms to cells. Similar considerations will apply to
research activities that are aimed at coarser levels of spatial
resolution. With current Compute Canada reso
urces leading
researchers are working with costly quantum mechanical (QM)
methods that have high accuracy but are limited to a maximum of
1000 atoms. On
-
the
-
fly molecular dynamics are typically limited to
some tens of picoseconds, even less for the larger

systems. In the
world of atomistic Molecular Dynamics (MD), the quantum
mechanical behaviour is replaced by a Molecular Mechanical (MM)
force field which allows many aspects of the dynamics and statistical
mechanics to be explored for much larger systems

(100 000 atoms)
for longer simulation times (the state
-
of
-
the
-
art is approaching a
millisecond). Coarse
-
grained approaches, in which a number of
atoms are frozen together into a combined particle whose external
dynamics is then studied, will increase the
se limits to perhaps a
million atoms and simulation times of the order of seconds. These
classical MM simulations do not allow the possibility of treating
chemical reactions so one of the current paradigms combines QM
and MM approaches in so
-
called QM/MM m
ethods. A typical
example would follow a chemical reaction taking place in the central
QM part of a protein with a model of around 100 atoms in the active
site, embedded in an MM model of the surrounding protein and
water of 100 000 atoms. Current effort
s focus on determining
appropriate methods for following the free energy of the reaction
with good statistical sampling of the motion of the environment.

Objective with Tier 1

Having substantial allocations on a tier
-
1 (peta
-
scale) computer
would allow ma
ny of these applications to be taken to the next level,
towards a true integration of time and length scales. It would, for
example, be possible to follow a chemical reaction with significant

48

|
P a g e


conformational changes and important fluctuations. Examples th
at
are currently being formulated and that would be ready for peta
-
scale facilities include the study of transcription by the RNA
-
polymerase enzyme, a drug
-
delivery model that involves siRNA in a
lipid mixture with water/salt and nano
-
pores that are being
considered for DNA sequencing devices. The proof of principle for
simulations on the atoms
-
to
-
cells scale should be feasible on a peta
-
scale machine, preparing the ground for the exa
-
scale which will
extend the models to tissues and organs.

Potential Appl
ication

While most of the frontier work is of a fundamental nature, aimed at
understanding the processes of life, there are already applications in
drug design and in biomedical engineering devices, for example.




Research Area: Health and Related Life S
ciences and Technologies

Research Objective

An important and fast growing area is the

bioinformatics applications
in biology and medicine. These include

applications dealing with the
analysis of vast amounts of complex data on

genome sequences,
gene expres
sion, epigenetic modifications, and genome
-
wide
association studies. The areas of proteomics, which involves

large
-
scale analyses of data of on protein expression and protein
-
interactions, or metabolimics

referring to comprehensive profiling of
metabolite
s (small molecules

processed by the cell), are also rapidly
changing the landscape of both fundamental

and applied health
research.

Objective with current resources

Address some of the pressing needs of
current
genomes
-
wide
sequencing technologies (Next

Generation Sequencing), which
routinely produce terabytes of sequence data that need to be stored
safely and processed in close to

real time. Adapt HPC and networking
environments to deal with the analysis and integration of large
heterogeneous Datasets a
nd the privacy requirements that might be
associated.

Objective with Tier 1

Faster and more comprehensive analysis of increasing
ly

large and
complex data sets. The time from data to knowledge and therefore
medical advances would be reduced by several order
s of magnitude.

Potential Application

Disease prevention,
diagnostics and treatment, personalized
medicine
.









49

|
P a g e


Research Area:
Communications and Information Technology

Research Objective

While silicon is the material backbone of cu
rrent communications
and information technology, advanced materials are constantly
being developed and studied for better performance either in
processing, switching or for displays. One can also envisage entirely
new technologies based on molecular elect
ronics or on quantum
computers. In all cases, accurate ways of simulating material
properties will accelerate development. Examples of materials under
study include graphene, various organic conductors and even
superconductors.

Objective with current reso
urces

Current resources allow first principles calculation of electronic
properties of materials for elements containing valence electrons in
the s or p shells in the first few rows of the periodic table. Commonly
used programs include Wein2k, Abinit, Gaus
sian etc… By contrast, in
the case where electron
-
electron interactions are important, either
in narrow bandwidth organic materials or for elements containing d
-
shell electrons, as one often finds in unconventional
superconductors, the difficulty of the ca
lculation increases
considerably. Only highly simplified models can be studied. Current
resources can combine first
-
principles calculations with methods
that take into account strong electron
-
electron interactions but only
at a rudimentary level. For examp
le, in the new pnictide
superconductors, such methods can handle only the normal state,
not the superconducting one.

Objective with Tier 1

To perform more realistic atomistic modeling and to take into
account the effects of disorder and of nano
-
scale inh
omogeneities
that are important in practice, increases by orders of magnitude in
computing power are necessary. As an example of what has already
been achieved with about 50,000 cores, a new algorithm to enable
400+ TFlop/s sustained performance in simulat
ions of disorder
effects in high
-
Tc superconductors (G. Alvarez, et al.

SC 2008, IEEE
Press, Piscataway, NJ USA 2008) has won the
Gordon Bell Prizes


awarded by the
Association
for Computing Machinery

in conjunction
with the
Institute of Electrical and Electronics Engineers

each year at
the
Supercomputing Conference

to recognize outstanding
achievement in high
-
performance computing applications. High
-
temperature superconductors are considered grand
-
challenge
problems but other new materials with interesting electronic
pr
operties pose the same challenge.

Potential Commercial
Application

Patents on materials that form the basis of new electronic devices
for the communication infrastructure could lead the development of
entirely new industries.


Even software companies coul
d emerge to
develop and sell the successors of Wien2k, Gaussian and the like. For
example, the company Atomistix
which develops software
solution for nanoscale electronic device modeling was founded
in 2003 based on such research at McGill University

(See

50

|
P a g e


http://www.quantumwise.com/

)