Cloud Computing (COMM034) Week 2 PaaS Dr Lee Gillam l ...

meatcologneInternet και Εφαρμογές Web

3 Νοε 2013 (πριν από 3 χρόνια και 9 μήνες)

95 εμφανίσεις

1

Want Sustainable and Green?

Seed Computing

Clouds



Dr Lee Gillam


l.gillam@surrey.ac.uk

What is our scope?


Green IT

and
Sustainability

often used interchangeably:


Typically, with focus on natural resources


more compute power
consumes more materials and energy;


Waste products throughout the product lifecycle, and familiar
(Greenpeace) images including those of children burning plastic
wires to obtain valuable copper.


Highly toxic elements in laptops and mobile phones.


But we probably also need to consider the sustainability of
IT, particularly in the UK.


Fewer “IT” graduates, but can technology reduce numbers needed?


Graduates need to know the business impacts of such changes


Cloud Computing demonstrates a lot of what we can be
thinking about.


So, just by seeding a few Clouds
….

Sustainability and Green IT?

3

A pile of electronic waste
on a
roadside

in Guiyu, China

Electronic waste in Guangdong, China As much as
4,000 tonnes of toxic e
-
waste are discarded every hour.
Vast amounts are routinely and often illegally shipped
as waste from Europe, USA and Japan to places where
unprotected workers recover parts and materials
.

Sustainability of UK IT?


Number of HE students and graduates at a record high but
HE Computing
student numbers and graduations are falling.


Students:
137,650 (2003
-
4); 106,910

(2006
-
7)


Graduations: 37,445 (2004
-
5) down to 31,270 (2006
-
7)


IT labour market seen growing.


Continuing demand for staff in the IT labour market leading to a net growth in the
number employed.


Single largest area of growth: Software Professionals, a role for which a high
degree of technical knowledge, capability and training is required.


Decline in computing graduates results in fewer “new entrants” from HE with
necessary deep
-
based technical skills.

Source: CPHC report: A study on the
IT labour market in the UK.(2008)

Structure of Talk


Introducing Cloud Computing


Software as a Service (SaaS)


Platform as a Service (PaaS)


Infrastructure as a Service (IaaS)


Concluding Remarks


So, what are these Clouds?


Already an environmental concern:



Make IT Green: Cloud Computing
and its Contribution to Climate
Change


http://www.greenpeace.org/raw/content
/usa/press
-
center/reports4/make
-
it
-
green
-
cloud
-
computing.pdf

-


12 pages,
7MB
.

Spot the Cloud

What is the Cloud?

What is the Cloud?

What is the Cloud?


Microsoft’s $500 million Chicago data center,
one of the
largest
data centers ever built,

spanning more than 700,000
square feet (Man Utd pitch about 80,000 sq.ft).


How might Clouds be built?


See Microsoft’s video at:
http://www.microsoft.com/showcase/en/us/details/36db4da6
-
8777
-
431e
-
aefb
-
316ccbb63e4e


Environmentally
-
friendly Clouds

12


In June 2007 Google completed a 1.6MW solar installation
-

the largest U.S. corporate installation at that time. 9,212 solar
panels cover the rooftops of eight buildings and two solar
carports at the Googleplex.


Produces enough electricity to power 30% of Google's peak
electricity demand


http://www.google.com/corporate/solarpanels/home



Power usage effectiveness (PUE), which gives a ratio of
total energy in to energy used for compute, does not
appear to incorporate how “green” the energy is.

Environmentally
-
friendly Clouds

13

Activity

Google searches

CO2 emissions of an average
daily
newspaper

(100% recycled paper)

850

A glass of
orange juice

1,050

One load of dishes in an
EnergyStar

dishwasher

5,100

A five mile trip in the
average U.S.
automobile

10,000

A
cheeseburger

15,000

Electricity consumed by the
average US
household

in one month

3,100,000



http://www.google.com/corporate/green/datacenters/

What is the Cloud?


It’s
not just
data centres, but “reselling” capacity in data centres is
key:


Advantage of economies of scale
to reduce the cost
-
per
-
unit service
provision and maximize the opportunities for efficient energy use in power
and cooling systems


machine room vs. broom cupboard?


Remove redundant server components
during production (e.g. sound
cards, USB ports).


Usage can be maximized



salesforce.com serves more than 1.5M users
(and 55,000 enterprise customers) every day with less than a 1,000 servers”
(500 used to mirror data).



“As of January 31, 2009, … a single, third
-
party Web hosting facility located on
the west coast of the United States, leased from Equinix, Inc. …. replicated in
near real
-
time in a separate [Equinix] back
-
up facility located on the east”.


What infrastructure duplication would occur if each enterprise
customer had their own in
-
house system?

http://www.slideshare.net/AmazonWebServices/london
-
aws
-
exec
-
wv
, slide 16.

Traditional IT


unanticipated

success can be disastrous: can’t scale to meet demand.

http://www.slideshare.net/kvjacksn/aws
-
2008
-
cloud
-
computing
-
ncoic
-
presentation
, Slide 27

“Cloud” PUEs

Source: Make IT Green: Cloud Computing and its Contribution to Climate Change

Where is the Cloud?


Amazon offers use of data centres in four availability zones
(US East/West, Ireland, Singapore).


Amazon.com also uses other data centres.



In 2003, Amazon began to use Equinix as a data centre provider.
Equinix has
50 data centres in 19 locations worldwide
,
with 5 data centres, and the company’s 50
th
, in and around
London.


The
Salesforce

CRM runs in Equinix facilities; further Equinix
customers include (or have included) Myspace.com, NASA, Fox
Sports Interactive Media and Sandisk.

Where is the Cloud?


Facebook is building its first data centre, scheduled for completion
in 2011, in Prineville, Oregon with an estimated cost around
$200m


Facebook has used Digital Realty Trust, who also provide a data centre
facility to
Terremark

and report some 50 Fortune 500 customers, and DFT
data centres, and has been alongside Twitter in Fortune Data Centers.


Terremark

own and operate NAP of the Americas


Miami, a facility that
houses, amongst others,
NewServers
.


ElasticHosts

use Peer 1 and BlueSquare Data. Peer 1 has data
centres in 16 locations, 14 in America/Canada, a London data
centre, and a mainland European data centre in Amsterdam.


NewServers
,
Terremark

and
ElasticHosts

are all “Cloud”
providers (more about them later also).

What is the Cloud?


“Reselling” capacity in data centres.


Buyer beware:
For Cloud Computing users, care needs to be
taken in selecting multiple Cloud providers in the hope of
constructing redundant systems. It is possible that the physical
infrastructures that underlie some of these systems are highly
co
-
located.


Cloud Storage solutions, JungleDisk and ElephantDrive, are both
reselling Amazon storage (S3) with bespoke functionality at slightly higher
prices, perhaps with some “free” space.


Need to know data centre location and service provider to ensure
adherence e.g. to the Data Protection Act 1998, additional regulatory
requirements,
AND because I may think I’m building
redundancy in when I’m not.

Business sustainability?


Some organizations need never purchase a server, own
software licenses, or worry (so much) about


whether current on server software licenses?


when to schedule next software upgrade?


what if the hardware fails at inconvenient times?


how do I manage technology assets?


what to do with old hardware?


how to treat depreciation of IT assets?


when can I afford to add capacity?

21

Traditional vs Cloud

Traditional Systems

Cloud Systems

File server

Google Docs

MS Outlook, Apple Mail

Gmail
, Yahoo!, MSN

SAP CRM/Oracle CRM/Siebel

SalesForce.com

Quicken/Oracle Financials

Intacct/NetSuite

Microsoft Office/Lotus Notes

Google Apps

Off
-
site backup

Amazon S3

Server, racks, and firewall

Amazon EC2
,
GoGrid
,
Mosso

22

NB: Just a sample!



from Reese, Table
1
-
1

SaaS

SaaS


SaaS probably the most mature *aaS of Cloud Computing


For some, SaaS began with mainframes!


Software as a package, with license, distributed on media,
installed by individuals / corporates


costs all round
.


Software becomes a download installed by individuals /
corporates


cheaper for producer
.


Software becomes large, difficult to install and configure, and
needs new hardware


more expensive for consumer
.


Company costs increased by software “bloat”; updates become
expensive to install, test, deploy, remove (on failure) etc.


Software usable over the internet.


24

SaaS


Advantages?


Bespoke infrastructure not required


Reduced system administrator loading


No upfront software costs


Users supported by the organisation that knows the product and how to
make it work well


BUT, some setup costs necessary. Training required, but system consistency
preferred, so incremental changes may reduce training overhead.


Disadvantages?


Costs may be “per seat” rather than “per use”.


Always
-
on connectivity / “thin pipes” problem.


Vendor lock
-
in?


Trust in third party?


Limited support in Service Level Agreements?


Emotional attachment to physical systems?


25

SaaS: Google Mail / Apps


Trinity College Dublin


Legacy

email system that lacked the features that students
wanted; Achieved
cost savings
associated with labour and
operational efficiencies


University of Westminster


Reported £1 million savings on IT costs


Universities of Portsmouth, Sunderland, Loughborough...

http://www.google.com/a/help/intl/en/edu/customers.html



Los Angeles City Council (2
nd

largest US city) outsources e
-
mail to Google;
$7.25
-
million contract
(5 years) to move
all 30,000 city employees.

26

Other SaaS

Application

Examples

Calendar

Google, Yahoo, Windows

Live,
CalendarHub
, Hunt Calendars, Calendar Net

Schedules

Diarised,

Windows Live Events,
Schedulebook
,
AppointmentQuest

Planning / Task

Management

Bla
-
bla

List,
Hiveminder
, Remember

the Milk,
Tudu

List,
HiTask
,
Zoho

Planner

Event Management

Conference.com,
RegOnline
, Event Wax

Project Management

BaseCamp
,

Project Drive,
Zoho

Projects,
onProject

Web Databases

Zoho

Creator /
Zoho

DB & Reports,
MyWebDB
,
Cebase
,
QuickBase
,
Lazybase

Bookmarking

BlinkList
,

Clipmarks
, del.icio.us,
Tagseasy

Photo

Editing

FotoFlexer
,
Preloadr
,
Snipshot

Photo Sharing

dotPhoto
,
Flickr
,

Photobucket
, Picasa Web Albums

Desktops

ajaxWindows
,
eyeOS
, g.ho.st,
YouOS

Web Conferencing

Genesys

Meeting
Center
, IBM Lotus

Sametime
, Microsoft Office Live Meeting,
WebEx,
Zoho

Meeting

Groupware

Contact Office, Google Sites, Project

Spaces,
teamspace

Blogs and Wikis

Blogger,
TypePad
,
WordPress
,

Pbwiki
, wikihost.org,
Wikispaces
,
Zoho

Wiki

27

PaaS

Google App Engine


Google has lots of data centres ....

From: Using Google App Engine

Where is my data??

Google App Engine


Google keeps buying computers.


Add or adjust network and computing resources to meet the demand.


Even Google may not be able to track what is running and where in
real time!


Systems must continue running, as far as possible, so independence of the
physical system vital.


Google has a software framework

an abstraction layer

that hides detail
about where data is located or which software is running on which server in
which data centre.


Abstraction layer provides flexibility in terms of dynamic reallocation of
resources to meet changing needs and demands.


Data, software, and computation can “follow the sun”


Where people are sleeping, data centres can undertake batch work such as Web
crawling, building search indexes, backups, maintenance updates, or supporting
load balancing for systems where the sun is.


Google’s
competitive advantage

here is cost
-
efficient scaling

Adapted from: Using Google App Engine

GAppE: What is it?


So, if you have a good software framework running in your
data centres, why not let others use it?


Applications run in a
sandbox.


Storage is Google’s Datastore (abstracted away from physical
storage)


Can retrieve URLs, but not open direct network connections


Applications are monitored


performance can be catered for;
bad applications can be stopped.


At specific levels of use (application popularity), charges apply

Adapted from: Using Google App Engine

GAppE: How does it work?


Google cloud like a mobile phone network:


Programs and data “roam”


your software could be anywhere in the
world at any given time.


Web requests (calls) find your software, regardless of where your
software happens to be running.


IP address of your Google application may be different, depending on
where you hook in.


Google’s system determines which data
center
(s) can run your application or
perhaps are already running your application.


Proximity, loading, or storage associations (where the data is “at rest” at the
moment) may be factors.


If your application experiences a sudden spike of traffic in the United
Kingdom, Google will likely copy your program and some of your data to
one of its data
centers

there, start your application in that data
center
, and
partition requests.


Scaling just happens!

Adapted from: Using Google App Engine

GAppE: How does it work?


Popular applications may be running in parallel in a number
of different data centers.


Unpopular applications may not be running until requested.


You don’t know where your application is, or if it is running.


Google hides all these details from you.


Users don’t really notice.


“Somehow, it should just work”


Adapted from: Using Google App Engine

GAppE: What else does it do?


When you run on your own web servers (“heavy lifting”):


Which O/S, version, patchlevel? (How many?)


How patched? (if Windows, reboot??)


Antivirus? Firewall?
(D)DoS


have to contact Google; may cost you.


Which DB, version, patchlevel?


DB on same machine as web server?


DB across multiple machines? DB as a bottleneck?


Support for peak demand? Unexpected demand? (When?)


When is an upgrade to any kind of capacity needed?


Under GAppE, Google’s problems!


Of course, costs start to apply (but economics need to be
considered anyhow).


Can you do it cheaper than Google can? Can another
provider …?

IaaS

IaaS


The cloud competes against two approaches to IT:


Internal IT infrastructure and support


ownership, no matter where it is located


pay staff


buy replacements (or cannibalise existing machines)


requires
Cap
ital
Ex
penditure (CapEx)


Outsourcing to managed services


rental (fees


Op
erational
Ex
penditure, OpEx)
-

someone else owns your
servers and keeps them running


rental company pays staff


managing infrastructure is their problem
-

replacement depends on the
service
-
level agreement
(SLA).

Adapted from Reese, Ch1.

Amazon Web Services (AWS)


EC2: APIs for
provisioning
,
managing
, and
deprovisioning

virtual servers inside the Amazon cloud.


Any application anywhere on the Internet can launch a virtual
server in the Amazon cloud with a single web services call.


Several “data centres” (availability zones)


Amazon’s EC2 U.S. footprint spans three (
or more
) data
centers on the East Coast of the U.S. and two (
or more
) in
Western Europe.


You cannot mix and match U.S. and European environments,
though you can run traffic between them
.

Adapted from Reese, Ch1.

Amazon Web Services (AWS)


Servers run a highly customized version of the Open Source Xen
hypervisor using
paravirtualization
.


provides isolated computing environment for guest servers.


Guest servers are set up as Amazon machine images (AMI) with an
operating system and a set of software; you will need a defined
software stack
for the functionality that you require.


Three kinds of storage directly relevant to EC2:


Ephemeral

storage


Block

storage (SAN
-
like) and persistent across time


Simple

Storage Service (S3)


can be used for cloud
-
based persistent
storage; used for AMI storage


AMI staged to disk from S3 (KVM)


Competitors may also provide persistent internal storage for
nodes to make them operate more like a traditional data center.

Amazon Web Services (AWS)


EBS and research data


move the analysis to the data....


Amazon Web Services (AWS)


EBS and research data


move the analysis to the data....


IaaS


Amazon Web Services (AWS)


including EC2, S3, ….


Rackspace


Flexiscale


GoGrid


Joyent


Terremark vCloud Express


ElasticHosts


NewServers



Why buy another machine??

Concluding Remarks

Buyer Beware


The market is full of supposed “Cloud” providers


but it might
not be “Cloud” as you expect it or if you subscribe to
NIST/Gartner (etc.) definitions or Open Cloud Manifesto.


A variety of pricing structures exists.


Everybody has a faster Cloud than everybody else.


Exercise for the reader


try and figure out what the Cloud
offering is from these providers.


Carpathia Hosting Inc:
http://www.carpathiahosting.com/



Layered Technologies:
http://www.layeredtech.com/



3Tera:
http://www.3tera.com/



Skytap:
http://www.skytap.com/

New Clouds env. + ?


Google Patent Application: “A system includes a
floating

platform
-
mounted computer data center comprising a plurality of
computing units,
a sea
-
based
electrical generator in electrical
connection with the plurality of computing units, and one or more
sea
-
water cooling units
for providing cooling to the plurality
of computing units.” .


An “offshore” data centre


but how far off shore?! (Tax avoidance vs
legislative difficulties?)


New Clouds env. + ?


Some combination of:


on the site of a Nuclear Reactor?


at the poles to provide free cooling


or out in space to provide
“really” cool computers?


on oil rigs, using “flare gas” for energy?


with less
-
predictable provisioning out at sea to make use of
hydroelectric and wind power?


Environmentalists and legislators will be looking at this for
years to come!


Cloud implies various issues


Green


SaaS, PaaS, IaaS


V12n


Multi
-
tenant


Clusters, HPC, HTC, P2P, Grid,
Web services


Load balancing, metering and
monitoring, utilization


Exotic architectures


Distributed data


ACID?
Federation, Replication


Economics


CapEx and OpEx,
billing and TCO


International Laws


Copyright,
DP/Privacy, Licences


Ethics?





“heavy lifting”


gone


Availability, segmentation,
redundancy and backup (and
security)


Different problems emerge


data
-
centre centricity (where and which?)



Additional Reading


Reese, G. (2009) “Cloud Application Architectures: Building
Applications and Infrastructure in the Cloud”. O'Reilly Media,
Inc. ISBN(13): 978
-
0596156367, 204 pages.


Above the Clouds:

http://www.eecs.berkeley.edu/Pubs/TechRpts/2009/EECS
-
2009
-
28.pdf



Amazon AWS:

http://docs.amazonwebservices.com/AWSEC2/latest/DeveloperGuide/



Sun’s Introduction:

http://www.sun.com/featured
-
articles/CloudComputing.pdf



Cloud Computing and the Law:

http://www.law.ed.ac.uk/ahrc/gikii/docs3/mowbray.pdf


47

Shameless Plug

Nikolaos Antonopoulos and Lee Gillam (Eds.) (2010): Cloud
Computing: Principles, Systems and Applications. Springer.
ISBN: 978
-
1
-
84996
-
240
-
7.
Due: August 2, 2010


Thank You



Dr Lee Gillam

L.Gillam@surrey.ac.uk