Cloud - AIM Utah

basheddockSoftware and s/w Development

Feb 21, 2014 (3 years and 6 months ago)

87 views

CLOUD
COMPUTING
Victoria
Livschitz

CEO, Grid Dynamics

3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

1

CLOUD = ECONOMIC MODEL

3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

2

9/18/09

© 2009 Grid Dynamics


Proprietary and Confidential

3


Building Blocks


Hardware infrastructure (
IaaS
)


AWS,
GoGrid
,
Rackspace
,
Mosso
,
Flexiscale


Platform (
PaaS
)


Force.com, MS Azure, Google App Engine


Software (
SaaS
)


Salesforce.com
,
Gmail



Cloud principles


Pay
-
as
-
you
-
go


Self
-
service


Programmable


Dynamic provisioning


Cloud scope


Public, Private, Hybrid


3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

4

3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

5

Shared
instances of
applications

Clustered
computing

HPC
applications

Hosting
infrastructure

Distributed
computing

Web Hosting


Cloud platform services


Storage and Data Services


Scalable archiving and backup solution


Relational, Object or Column Databases


Network
-
attachable block storage


Messaging


Other: Auth, CDN, URI fetching, Task Scheduling, Image Processing, etc.


Cloud middleware


Grid Computing Middleware


Data grids


Provisioning systems


Cloud API


Monitoring


Management


Orchestration


3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

6

Cloud

API

Middleware

Platform
services


Benefits of scale


Large cloud provider can always invest more

into robust cloud services than you


Illusion of “infinite capacity” for processing and storage resources


Bursting: Employ large amount of servers with ease


Risk management


Most infrastructure risks moved to cloud provider

(HW and network failures is provider’s problem)


Resilient and redundant cloud services as building blocks


Opportunity to automate failover and failback


Speed of innovation


Low entry cost for innovative programs


Easy to try a lots of ideas and not meet IT
-
related bounds


3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

7


Raw performance


Virtualization tolls CPU and I/O


Data
-
compute affinity


Moving data in and out is costly


Firm SLA


Fair share of resources is not
guaranteed in all scenarios


Security and Data privacy


Cloud abstracts out where
actual data processing occurs


Digital Millennium Copyright Act
(DCMA) and
CryptoExport

laws


Lock
-
In


No standards so far




Opacity


Reliability and safety
technologies are proprietary and
unknown


Legacy as a Service


Rarely cloud
-
friendly


Deployment


Automate, automate, automate


Everything fails all

the time


Cloud provider outages, hidden
and obvious


Unexpected Cloud API and
Services evolution

3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

8


Business Agility


Internal IT is slow to respond to business change +


3
-
12 months to procure & deploy +


Cost of delivery often 10x the cost of hardware +

= incentive at BU level to leverage public cloud, rather then internal IT

= incentive for IT to adopt private cloud


On Demand Scalability


Some applications JUST NEED SCALE


100
VMs

for 1 hour = 10 VMS for 10 hours


Impossible is now easy and cost effective


Decrease in IT Complexity


Less compelling today due to longer term to benefits, security concerns,
tradition etc.

3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

9


CLOUD = BUSINESS AGILITY


3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

10


Originally: a “must have” website extension of Macys
marketing arm


Today: a premier shopping destination, generating $1B+
in business


Growing at 20%+ per year, vs. 3
-
5% traditional bricks &
mortar retail


Inherited monolithic, legacy infrastructure from

corporate IT


How to compete with agile, pure play .
coms

like Amazon?

3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

11


6
-
12 static staging environments


Continuous feature development


Frequent push
-
to
-
production cycles


Multiple branches for special sales programs


QA provides environment configuration specs


sets up
the environment for specific date & build


1
-
2 days latency to allocate, set up and configure physical
QA environment


Availability of resources in operations


Specs improperly communicated


several iterations to set up
environment

3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

12

3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

13

3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

14


Time to deploy QA
environment

~30
-
60
mins


Full isolation of QA
environments


One
-
click deployment
of several
environments in
parallel


Instant disposal of
environments


Cost savings:

1 QA environment =
$0.76/hour

3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

15

3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

16

MST

RackSpace


GoGrid


Equipment

$22,000

$53,554

$177,192

Storage

$1,000

$5,400

$5,400

Admin Cost

$15,600

N/A

N/A

Infrastructure Cost

$60,000

3
-
year Total

$98,600

$58,954

$182,592

Utilization Adjustment

$0

-
$47,163

-
$146,074

Total Cost

$98,600

$11,791

$36,518

Standard MCOM Environment


Small Server

2

Larger Server

2

Storage (TB)

1

Utilization

20
%


CLOUD = WIDER MARKETS FOR
LEGACY PRODUCTS


3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

17


World leader in testing and

measurement equipment


16,000 employees.

$4.5 billion revenue


Portfolio of proprietary, niche software products,
complimentary to core business


Legacy architecture, on
-
premise deployment:


Time to deploy


months


Cost to deploy


hundreds of thousands

Challenge:
How to compete with agile, pure play
SaaS

offerings for SMB market share?


3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

18


Reuse application


Application is a black
box


No porting of code



Deploy to public cloud


Time to deploy


minutes


Cost to deploy


virtually none


Fit to
SaaS


Cloud API integration


Instance management


User access
management


Usage metering &
billing

3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

19

3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

20

Analytical Applications

Monte
-
Carlo

Excel Analytics

Batch Processing


CLOUD = SCALABILITY ON
-
DEMAND


3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

21

TRADITIONAL APPROACH

NEW APPROACH

Always comput
e intensive, sometimes data intensive

Runs

as a batch

Runs as a service

Fixed

static footprint

Use idle compute cycles (CPU
scavenging)


Dynamically

scalable

Based on popular scheduler
-
based
grid frameworks

Based on emerging HPC technologies

Not designed for near real
time processing

Oriented to near real time
processing


Major investment bank


Annuity calculator application


Monte
-
Carlo simulation with geometric Brownian motion (GBM)


Fully parallelizable algorithm


Customer talks to an agent and agent gets back to the customer next
business day


Currently nightly batch job computes the annuity amounts


Problems with current approach


System is constrained by time available for batch


Customer satisfaction can be improved if this can be computed on spot, in
near real time


Adding new resources to system is hard and expensive

3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

22


Launch and shutdown the application on demand


Scale up or down based on the size of the problem


Complete the simulation in near real
-
time


Model functionality should be reusable


Security


Re
-
use existing Monte Carlo models (written in C++)

3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

23


Business service can be
launched or dismissed on
click of a button


App is up and serving in
less than 4 min


Scale up under 2 min


~1M simulation draws in
near real time


Fast and secure: app code
and reference data boot
up from a local storage
cloud


3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

24

FINALLY, SOMETHING A BIT OFF
THE MAIN PATH:


PARALLEL EXCEL ON THE CLOUD

3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

25


Ubiquitous tool


Financial analysts think in Excel


Excel + VBA is current financial analyst IDE


For many financial institutions, Excel is a main data analysis tool


Used by analysts and engineers


Limited Programming Model


Single threaded, memory limited, not that performing


Need to Run Large Excel Workloads


Parallelization of workload and data is the only way out


On
-
demand infrastructure to run parallel excel

3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

26


Calculation Flow


DAG of calculation units (Macro,
UDF, Workbook
recalc
)


Representable

as “DAG table” or
task dependency table


Data flow


Workbook as a system of
records and data
synchronization point


Moving around workbooks is
costly


moving data deltas is
essential


Template regions are used to
capture input and output
parameters





3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

27

3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

28

Scheduler

Compute Nodes

(MS Windows & Excel)

Staging Server

Cloud (Private or Public)

Private Link

Or

Internet

Customer Premises

HTTP or FTP Server

(Only for Public
Clouds)

User PCs

(MS Windows & Excel)

Web Server

1. Submit Job

2. Stage Workbook In

4. Stage Result Out


Enterprise adoption of public cloud today is

characterized by:


Use cases around non
-
mission
-
critical systems


Limited to spot scenarios that generally fall into two buckets: a) business
agility through
quick deployment
;
b
)
bursty

scale
-
out

of compute intensive
apps


Public cloud competes with internal IT and accelerates
private cloud adoption


Private cloud adoption model:


Consolidation


Virtualization


Automation


Utility


Market

…few IT organizations reached the automation milestone…


3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

29

© 2010 Grid Dynamics


Scaling Mission Critical Systems

3/11/10

30


Grid Dynamics a global leader in delivering
Scalable

and

Elastic

application infrastructure solutions for mission
-
critical business functions.


Founded in 2006


Top
-
20 fastest growing private US company (according to
500, Inc.)


200 Engineers

3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

31

3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

32

Extreme transaction processing,
next
-
generation distributed data
stores, enterprise messaging,

Retail, billing

Search, imaging, analytics

Trading, simulations

HPC, multi
-
core, SMP

“Take me to the cloud”

“Help me build a cloud”


Private, hybrid and public

HPC (WCCS)

Cloud (Azure)

IMDG (Velocity)

3/11/10

© 2010 Grid Dynamics


Scaling Mission Critical Systems

33