Why Exadata wins - real exadata case studies from Proact portfolio - Fabien de Visser and Frits...

Preview:

DESCRIPTION

Presentation from conference "Oracle Day 2011" in Estonia 11.03.2011 Nordic Hotel Forum

Citation preview

Oracle Exadata

Fabien de Visser

Enterprise Account Manager

PROACT

Frits Hoogland

Oracle ACE Director

VX Company

PROACT Netherlands

Employees: 80

Revenue: E 35 million

March 11, 11

Focus: Storage, backup, recovery, archiving

VX Company

Employees: 300

Revenue: E 70 million

Focus: Oracle, Microsoft, Java

Consulting & projects

Case Study

Case Study: Erasmus University

Case: store 100 human genomes...

...for the first year.

Grow further in the next years.

Current problems

Too much data, and growing

Taking too much time to diagnose

� Data is a sample DNA variations set

� Table: CG_VAR– Size: 133’425’004’544

– Extents: 2’228

– Blocks: 16’287’232– Blocks: 16’287’232

� No indexes

� No constraints

� Tests are done on VX Company & PROACT half rack Exadata V2

What is done?

In all cases the following simple SQL is executed:

SQL> select count(*) from cg_var;SQL> select count(*) from cg_var;

� Response time: 695 seconds– All 133GB is read and send to upper layer

Nr Exadata features

Parallel Disk type

1 - Serial HDD

– All 133GB is read and send to upper layer

– 127,238 IO’s

– Time profile:

– Restriction: Disk (652/127238=0.005)

direct path read 652

DB CPU 45

total time 695

� Response time: 403 seconds– All 133GB is read and send to upper layer

Nr Exadata features

Parallel Disk type

2 - Serial FDD

– All 133GB is read and send to upper layer

– 127,238 IO’s

– Time profile:

– Restriction: Disk (359/127238=0.002)

direct path read 359

DB CPU 44

total time 403

� Response time: 18 seconds– All 133GB is read and send to upper layer

Nr Exadata features

Parallel Disk type

3 - 64 HDD

– All 133GB is read and send to upper layer

– Time profile:

– Restriction: Disk- 133GB/4=33GB/18=1.84GB/s/n ; 4GbFC=33/0.4=82.5s

direct path read 256

DB CPU 16

total time 272

� Response time: 13 seconds– All 133GB is read and send to upper layer

Nr Exadata features

Parallel Disk type

4 - 64 FDD

– All 133GB is read and send to upper layer

– Time profile:

– Restriction: Disk

direct path read 182

DB CPU 18

total time 200

� Response time: 45 seconds – All 133GB is read and 19 GB send to upper layer

Nr Exadata features

Parallel Disk type

5 SS Serial HDD

– All 133GB is read and 19 GB send to upper layer

– Time profile:

– Restriction: CPU

cell smart table scan 8

DB CPU 37

total time 45

� Response time: 41 seconds – All 133GB is read and 19 GB send to upper layer

Nr Exadata features

Parallel Disk type

6 SS Serial FDD

– All 133GB is read and 19 GB send to upper layer

– Time profile:

– Restriction: CPU

cell smart table scan 3

DB CPU 38

total time 41

� Response time: 13 seconds – All 133GB is read and 19 GB send to upper layer

Nr Exadata features

Parallel Disk type

7 SS 64 HDD

– All 133GB is read and 19 GB send to upper layer

– Time profile:

– Restriction: Disk- 133GB/7=19GB/13=1.5GB/s

cell smart table scan 168

DB CPU 14

total time 182

� Response time: 5 seconds – All 133GB is read and 19 GB send to upper layer

Nr Exadata features

Parallel Disk type

8 SS 64 FDD

– All 133GB is read and 19 GB send to upper layer

– Time profile:

– Restriction: Disk- 133GB/5=26.6GB/7=3.8GB/s

cell smart table scan 65

DB CPU 14

total time 79

� Response time: 1 second– EHCC Query compression: 133GB is reduced to 11GB

Nr Exadata features

Parallel Disk type

9 SS + EHCC 64 FDD

– EHCC Query compression: 133GB is reduced to 11GB

– All 11GB is read and 260MB send to upper layer

– Time profile:

– Restriction: CPU

cell smart table scan 4

DB CPU 10

total time 12

Case study

9,300 Retail stores

98000 Employees

Based in Hong Kong

In Netherlands brands ‘Kruidvat’ and ‘Trekpleister’ are

used.

Case study AS Watson

Investigation into user satisfaction

BI performance issues

Better reports needed

Project started for new DWH

Performance 10x faster

Case study AS Watson

POC Netezza

Achieved 10 times performance improvement

But change in platform results in change of technique

POC IBM

Hardware: large P7 with huge storage system

Expected 30-40 times performance improvement

In real life achieved 15 times improvement

POC Oracle Exadata

On exadata quarter rack

Achieved 10-200 times improvement

Case study AS Watson

AS Watson decided not to go ahead with IBM

Netezza and Exadata remained

Netezza

Pro’s

• Netezza helps with importing data

• Fast

Con’s

• Again a new vendor

• No knowledge in organisation

• No OLTP

Case study AS Watson

Exadata

Faster than Netezza in general

Pro’s

• OLTP possible

• Possibility to consolidate

• Fits in existing Sun environment

• Already user of Oracle database software

Con’s

• Upgrade to newest version needed

• Extra knowledge of Oracle database software needed

Case study AS Watson

Conclusion

Price difference 20k

Extra consultancy offered

Order received for 300k, system already running in UK

Recommended