Noel Yuhanna Senior Analyst Forrester Research

ForrTel:Scaling Your Enterprise DatabaseNoel Yuhanna

Senior Analyst

Forrester Research

November 29, 2004. Call in at 12:55 p.m. Eastern Time

Theme

Scaling databases for high-end deployments is still

challenging, therefore careful planning and

execution remain essential

Agenda

• Database classification

• Largest OLTP/DW deployments

• High-end deployment challenges

• DBMS product scalability

• TPC benchmarks

• Scale-out vs. scale-up

• Best practices/recommendations

Forrester survey

What is your largest production enterprise database?

Less than 50gb

50gb to 100gb

100gb to 250gb

250gb to 500gb500gb to 1tb

Over 1tb 28%8%

10%

10%

23%21%

Survey Data: 52 Enterprises

Forrester survey

Percentage change in number of databases over the next two years?

Decrease

<10

25-50

50-7575-100

10-25

10%

2%4%

17%

19%

48%


Enterprise database classification

# of objects 100 or less 100-1,000 1,000-5,000 Over 5,000

DB size 10GB or less 10GB-250GB 250GB-1TB Over 1TB

Users supported 100 or less 100-1,000 1,000-10,000 Over 10,000

# of trans (min) 500 or less 500-5,000 5,000-10,000 Over 10,000

Small Medium Large Complex

Source: Forrester Research, Inc.

Largest OLTP databases

• Land Registry — 18TB on DB2 for z/OS

• British Telecom — 12TB on CA-IDMS

• UPS — 9TB — DB2 for z/OS

• US Patent & Trademark Office — 5TB Oracle

• Verizon Communications — 5TB SQL Server

Source: Winter Corp Survey, 2004

Largest data warehouses

• France Telecom — 30TB Oracle

• SBC — 25TB on Teradata

• Amazon — 13TB on Oracle

• Kmart — 12TB on Teradata

• Health Insurance Agency — 12TB on Sybase

2,000 sites running over

1 terabyte DB in production

Source: Forrester Research, Inc

Trends — Large OLTP/DW deployments

Database Continue To Grow

50TB

100TB

30TB

50TB

0

20

40

60

80

100

120

2000 2001 2002 2003 2004 2005 2006 2007

Ter

abyt

e

OLTP

Data warehouse

Largest OLTP and data warehouse deployments


Forrester survey

What are your most challenging database administration tasks?

Performance/Tshoot26%

Patch/upgrade21%

Change management

14%

Planning 11%

Rep/sync 8%

Backup/rec 8%

Security issues6%

Resource issues

6%

Database activities that are most challenging


High-end deployment challenges

• Database challenges

» Backup/recovery

» Performance, tuning, and troubleshooting

» Upgrade/patches

• Platform and infrastructure challenges

» Choosing the right platform/infrastructure

• Migration challenges

Oracle DBMS scalability

• Oracle RAC — scalability/availability

• Has the most tuning options

• High scalability on UNIX/Linux platforms

• Logical partitioning of data — range, list, hash, comp

• Strong in mainframe and high-end performance/scalability

• Many large database deployments

• Scalable for large data warehouses

• Recent high TPC-C results — show DB2 scalable

DB2 DBMS scalability

SQL server DBMS scalability

• Strong small to medium size database deployments

• Over one terabyte databases are challenging

• Data warehousing scalability still weak

• SQL Server 2005 — likely to address scalability issues

Sybase DBMS scalability

• Strong data warehousing scalability

• Reasonable OLTP scalability on UNIX

• Not many terabyte-size DBMS deployments

DBMS scale-up vs. scale-out

Scale-up(horizontal)

Scale-out(vertical)

Small

Medium

Large

Medium-Complex1-4

1-4 1-4 1-4

8-12

4-8

Complex 12+

Lar

ge

SM

P S

erv

ers

Clustered nodes

Top TPC-C (OLTP) benchmarks S

cale

-up

Scale-out

DB2

Oracle

SQL Server

3.2

1.1

1.0

.80

.78.76

.70.68

DB2, Oracle

, SQL S

erver

Oracle

RAC

In Million TpmC

TPC-H (DW) benchmarks (1-10TB)

(QphH- in thousands)

Sca

le-u

p

Scale-out

DB2

Oracle

SQL Server

86.2

62.249.1

45.2

34.328.9

27

22.3

21.0

35.1

26.1

20.2

5.1

3.3

Oracle

Oracle + D

B2

Enterprise DBMS scalability — positioning


Small

Medium

Large

Complex

Ora

cle

DB

2

Syb

ase

SQ

L S

erve

r

Info

rmix

Ing

res

MyS

QL

Po

stg

res

ql

Scaling enterprise databases

DBMS

Storage

CPU

-Defrag-Tune

Archive

Right sizingTune Balanced H/W

Optimize code for performance

Optimize

Tools

Network

“Choosing the right hardware infrastructure is critical for high-end deployment.”

SAN,NAS

64-bit DBMS deployments on the rise

• Available on all platforms — M/F, Unix, Linux and Windows

• DBMS priced same as 32-bit editions

• Most large App’s can benefit, but not all

• Key benefits:

» Large buffer cache — hit ratio

» More concurrent users

» Pin smaller tables in memory

» Faster data scans/sorts

• Suitable for:

» Large e-Biz applications

» Thousands of users — OLTP applications

» Large D/W — supporting complex queries

Future scalability: grid computing

StandardsStack integration

ManagementArchitecture

Heterogeneousness

Still Evolving:

Future scalability: data grids

Data grids

Cache

Infrastructure farm

Applications

Best practices/recommendations:

• Planning remains important

» People, infrastructure, DBMS

» Testing and implementation

• Benchmark before deploying high-end workloads

» Useful for migrations

» Capacity planning

• Hardware infrastructure is very important

» CPU, memory, disks — need balancing

» Storage: SAN/NAS — Network Appliance/EMC

» Understand the hardware limitations

Best practices/recommendations (continued)

• Scale-up for OLTP scalability dominates

» Proven scalability

» Expect major drive towards scale-out in coming years

• Database size impacts performance » Archive data

» Defragmentation can help

• Optimize DBMS for performance

» Tune — SQL and DB

» Use tuning tools — BMC, CA, Quest

Noel Yuhanna

[email protected]

www.forrester.com

Thank you

Entire contents © 2004 Forrester Research, Inc. All rights reserved.

Client Choice ballots for Application, Development, and Infrastructure open

December 15.Make your opinion count and vote today.

To view Client Choice research and current ballots, click here:

http://www.forrester.com/find?N=101

Noel Yuhanna Senior Analyst Forrester Research

Documents