grid computing for financial services

32
© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Grid Computing for Financial Services Steve Conn, Sales Development Manager at Intel Yinal Ozkan, Financial Services Technology Leader at AWS April 14, 2016

Upload: amazon-web-services

Post on 15-Apr-2017

715 views

Category:

Technology


1 download

TRANSCRIPT

Page 1: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

Grid Computing for

Financial Services

Steve Conn, Sales Development Manager at Intel

Yinal Ozkan, Financial Services Technology Leader at AWS

April 14, 2016

Page 2: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

Motivations for Cloud in a Simulation-Driven World

Scalability and agility

Secure global collaboration

Enterprise data governance

Page 3: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

Agility Is Enabled by Global Scale

Over 1 million active customers across 190 countries

800+ government agencies

3,000+ educational institutions

12 regions

33 availability zones

54 edge locations

Page 4: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

HPC Today

Page 5: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

1965-2015: 50 Years of Moore’s Law

Page 6: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

The Rise of the Microprocessor

† Codename Knights Landing, not drawn to scale

28MTIMES FASTER

Intel® 4004 (1971)10,000nm, 2300 Transistors

92 KOPS

Intel® Xeon

Phi™(2015)†

14nm 3D Tri-Gate, >8B Transistors

3 TFLOPS (peak DP-F.P.)

6

Page 7: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

Supercomputing Pulls All HPC Along

Source: Top500.org

’90 ’95 ’00 ’05 ’10 ’15 ’20 ’25

0

10

1,000,000,000

10,000,000

100,000

1,000

Top Machine

500th Machine

Sum of All

Top 500 FLOPS >50%1 CAGR For Past Decade

#500 system on Top 500 in 2005:

~1 TFLOP

10 years later…

…One 2S Xeon server today:

~1 TFLOP

Capability Waterfalls

Source: Intel

Page 8: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

HPC Fuels Insight and Innovation

Astrophysics

ManufacturingEnergy

Security

FinancialLife Sciences

Weather

Source: www.hpcwire.com/2014/01/02/top-supercomputing-discoveries-2013/

Page 9: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

HPC Today: Two (Connected) Worlds

Supercomputing

The “rest” of HPC

Page 10: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

Intel Cloud Edition for Lustre Software

Innovate Faster. Experiment Risk-Free • Enterprises are adopting HPC clouds…

• To increase and augment legacy infrastructure

• Aggregates capacity and I/O bandwidth

• Brings Lustre performance to dynamic cloud resources

• Instant, pay-as-you-go access to scalable clustered

storage

• Available today from Amazon Web Services

• Support directly from Lustre experts at Intel

HPC

Big Data

Cloud

Page 11: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

Cloud Edition for Lustre Client Cluster &

Lustre Cluster Model

Page 12: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

Performance

• The Intel Lustre solution is a fast,

scalable storage platform positioned to

accelerate application performance,

even with complex workloads.

• Intel Cloud Edition for Lustre software is

an ideal foundation for SAS Grid

workloads that require fast, scalable, and

cost-effective storage. 0

500

1000

1500

2000

2500

3000

3500

4000

4 OSS 8 OSS 16 OSS

OSS Write / Read Scaling

Write MB/s Read MB/s

Page 13: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

Grid Computing on AWS

Page 14: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

Our Pricing Structure

On-Premise (or “Private Cloud”)

The New Way

Use only what you need,

using on-demand, reserved, or spot

Flexible

The Old Way

High upfront capital cost,

high cost of ongoing support

Inflexible

AWS lets you pay for the grid infrastructure you need, when

you need it.

Page 15: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

The Old Way – Low Utilization, High Costs

Typical service utilization rates are low due to need to deploy

for peak needs…

Time

Page 16: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

The Old Way – Managing Utilization with Grid

Time

Higher grid utilization rates result in hidden costs, such as

longer queue wait times and delayed results.

Page 17: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

The High Cost of Queues for Grid Computing

Conflicting goals

• Grid users seek fastest possible time-to-results

• Grid workloads are not steady-state

• IT support teams seek highest possible utilization

Result

• The job queue becomes the capacity buffer

• Job completion times are hard to predict

• Users are frustrated and run fewer jobs

• Innovation is throttled by IT resources

?

Page 18: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

The World According to Central IT

Grid utilization over one week

Page 19: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

The World According to Business Lines

Queue wait times over one week

Business

impact!

Page 20: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

The Cloud Way – Scalability When Needed

Scale higher to reduce time-to-results: shorter wait times,

greater agility, and faster innovation cycles.

Page 21: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

In a secure Virtual Private Cloud

Automation and Auto Scaling allows easier

grid management and monitoring

Page 22: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

AWS Enables Lower TCO than On-Premise

Utilization fundamentally higher in

AWS cloud

• Aggregating non-correlated

workloads, scale, spot market

Amazon specific hardware designs

• Amazon custom servers & network

gear

• Direct purchasing of disk, memory,

& CPU

AWS Immense scale

• New data centers built each year

• 12 global regions, 33availability

zones

• Volume purchasing with highly

automated, highly optimized supply

chainTraditional

Data Center

Virtualized

Data Center

UPFRONT

COSTS

VARIABLECOSTS

VARIABLE COSTS

AWS

UPFRONT

COSTS

UPFRONT

COSTS

VARIABLE COST

Cost savings from running

internal IT more efficiently

Cost savings from moving to a

public cloud provider

Page 23: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

Customer Spotlight: FINRA

Page 24: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

Customer Spotlight: Dow Jones International

• From over 40 data centers down to 6

• Planning to migrate 3000 apps by 2015

• Saving $100M over 3 Years

1. Evaluate infrastructure

costs & architecture

VS

2. Make business case 3. Enable decision to

move to the cloud

Page 25: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

Multiple Consumption Models

On-Demand

Pay for compute

capacity by the hour

with no long-term

commitments

For spiky workloads,

or to define needs

Reserved

Make a low, one-time

payment and receive

a significant discount

on the hourly charge

For committed

utilization

Spot

Bid for unused

capacity, charged at a

Spot Price which

fluctuates based on

supply and demand

For time-insensitive or

transient workloads

Spot is a

game-changer

for grid

computing.

Page 26: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

On

Reserved Instances

100%

On-demand

Time

Spot

Optimizing AWS Spending with RI, On-Demand,

Spot

Page 27: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

All-In on Spot: Advertising Technology Company

Minimal use of RI

and on-demand, for

persistent database

tiers.

Use-case is time-critical

bid-for-placement

advertising analytics.

Page 28: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

Spot is used by this customer for a wide range of use-cases, including

Hadoop analytics and test/dev. Customer has implemented automation to

reduce impact of Spot terminations.

Fully Optimized:

Enterprise SaaS Solutions Provider

Page 29: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

On

3-Year RI

100%

On-Demand $

Time

Spot $

1-Year RI

3-Year RI

1-Year RI 1-Year RI

1-Year RI 1-Year RI 1-Year RI

Price reductions

New instance types

Time

Cost

Optimization

Performance

Optimization

Reduced Spending Over Time for Increasing

Performance

Page 30: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

Proactive Cost Optimization with Trusted Advisor

Page 31: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

AWS is more cost-effective than on-premises environments in both the short-term and long-

term, and at scale, for both variable and steady-state workloads, while also delivering return on

agility and enabling new experiences.

In Summary

Page 32: Grid Computing for Financial Services

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

Thank you!

Steve Conn, Sales Development Manager at Intel

Yinal Ozkan, Financial Services Technology Leader at AWS