The Impact of Cloud Computing on Data Warehousing

Tasso Argyros , Co-Founder and CTOApril 15th, 2009

Topics

Aster Data Systems

Introduction to Data Warehousing

Impact of Cloud on Data Warehousing

Aster nCluster Cloud Edition

2 Confidential and proprietary. Copyright © 2009 Aster Data Systems

IntroductionAster Data Systems

Who is Aster Data Systems?

Relational database for data warehousingsoftware that runs on big clusters of cheap servers

Founded in 2005Mayank Bawa CEO [Stanford InfoLab]Mayank Bawa, CEO [Stanford InfoLab]

Tasso Argyros, CTO [Stanford DSG]

George Candea, Chief Scientist [Stanford ROC]George Candea, Chief Scientist [Stanford ROC]

Roots Investors Recognition

4 Confidential and proprietary. Copyright © 2008 Aster Data Systems

IntroductionData Warehousing

Enterprise Data Warehousing

Frontline Applications

Move &Batch Load

ReportsAnalysis

EnterpriseData 

Warehouse

OperationalData StoreRecord

Transform& Cleanse

SourceOLTP 

Database

Frontline Applications

WarehouseDatabase…Frontline 

Applications

Slide 6

Trends in Data Warehousing

Richnessof queriesq

Size of data

1. Mix of queries changes as more users are added2 U h k d t h ithi d

7 Confidential and proprietary. Copyright © 2009 Aster Data Systems

2. Usage has peaks and troughs within a day

Implications on Infrastructure

Compute and storage requirements are high & increasing• Big SMP and SAN deployments

Infrastructure footprint is large • Upgrades are expensive in time and effort

Provision fornow + 3 years

80% of initial cost is infrastructure costRequirements now

8 Confidential and proprietary. Copyright © 2009 Aster Data Systems

MySpace (2007-09): Actual Deployment

$7K Server $7K Server

20TB150 TB350 TBFrontline Applications

$7K Server

$7K Server

$7K Server

$7K Server

$7K Server

$7K Server

$7K Server $7K Server

$7K Server

$7K Server

$7K Server

$7K Server$7K Server

$7K Server

$7K Server

$7K Server

$7K Server

$7K Server

$7K Server

$7K Server

$7K Server

$7K Server

$7K Server

$7K Server

Slide 9

$7K Server $7K ServerReports Applications Analysts

Data Warehousing is now “Cloud-Friendly”

$7K Server $7K Server$7K Server

$7K Server

$7K Server

$7K Server

$7K Server

$7K ServerSMP$7K Server $7K Server

$7K Server

$7K Server

$7K Server

$7K Server$7K Server

$7K Server

$7K Server

$7K Server

$7K Server

$7K Server

$7K Server

$7K Server

$7K Server

$7K Server

$7K Server

$7K ServerSAN

Slide 10

$7K Server $7K Server

Cloud Computing Impact on Data Warehousing

Public and Private Clouds

BENEFITS• Pay only for what you use• Fast scale-up (or down)

ENTERPRISE CONCERNS• Privacy/security of data in a

shared infrastructureD t t f d Fast scale up (or down)

• Reduce admin overhead • Data transfer speeds over public Internet

Cloud VariantsCloud Variants

PUBLIC CLOUD• Example: Amazon EC2• Typical users

PRIVATE CLOUD• Owned by large enterprise IT groups• Centralized infrastructure for use • Typical users

Startups/developersEnterprise experimenters

across the company• Address enterprise concerns of

security and data transfer speeds

Aster nCluster Cloud Edition

Proven: ShareThis is largest cloud-based DW in world on AWS Proven: ShareThis is largest cloud based DW in world on AWS • (2.2TB, growing to 10-18TB by year-end)

Easiest on-demand scaling in the market

First host-vendor-neutral offering

13 Confidential and proprietary. Copyright © 2009 Aster Data Systems

1. Elastic Scalability

Live Queries

Add Capacity

• Single-click scale-out and scale-down with no downtimeSingle click scale out and scale down with no downtime• Automated incorporation and load balancing in minutes• Database available even while loading, backup, export, restore, scale-up, re-provision, fault recovery, p, p , y

14 Confidential and proprietary. Copyright © 2008 Aster Data Systems

2. “Always On” Availability

Worker1 Worker3 Worker4 Worker5Worker2

• Cloud units WILL FAIL • Online backup and restore• Online load and export

15 Confidential and proprietary. Copyright © 2008 Aster Data Systems

3. Hibernating Services

No Queries

• HIBERNATE data to cheaper storage Release Capacity

p g•Release cloud units when no usage•Revigorate on-demand

16 Confidential and proprietary. Copyright © 2008 Aster Data Systems

4. Managing Workloads

Query Set 1 Query Set 2

Clone Warehouse

• CLONE service (data + compute) to a new pool •Re-assemble pool when usage declines

17 Confidential and proprietary. Copyright © 2008 Aster Data Systems

SUMMARY: Data Warehousing in the Cloud

1 Port Product 2 Innovate Product1. Port ProductEnsure compatibility

Ensure performance

2. Innovate ProductLeverage “infinity”

Leverage service APIsEnsure performance

Ensure features

Leverage service APIs

Enable new features

18 Confidential and proprietary. Copyright © 2009 Aster Data Systems