You are on page 1of 15

Approach to Performance

A Conversation on Holistic Approach to Performance

Performance - Art, Not a Drill

Hardware Sizing

Cores per node

Memory per node

Storage per node

Dataset and Throughput

Clustering (Sample for Illustration Only)

Data Model

Data Profiling

Critical Data Attributes

Threshold

Big Entities

Inline Extension

Auditing (TAIL in AE, custom SQL in SE)

Historical Data, Max Active/Existing Values

Custom Solutions

Large Result Set, Custom UI

Coding Best Practices

SMART Inquiries

Inquiry Level, OTS

Granular Services, not complex

Asynchronous Workloads

SDP / Evergreening /Entity Management

MQ over JMS queue

Tune SQL Query, Use Prepared Statements

Probabilistic Search

What should I Do to Improve Performance?

MDM AE:

http://www.ibm.com/developerworks/data/library/te
charticle/dm-1010mdmservertuning1/index.html
http://www.ibm.com/developerworks/data/library/te
charticle/dm-1010mdmservertuning2/index.html
http://www-01.ibm.com/support/docview.wss?uid=
swg27046292

MDM SE:

http://www-01.ibm.com/support/docview.wss?uid=
swg27042420

What should I Do to Improve Performance?

Oracle
http://www.oracle.com/technetwork/database/manageabi
lity/db-perf-tuning-ow08-131582.pdf
DB2

http://www.redbooks.ibm.com/redbooks/pdfs/sg247073.
pdf
JVM Tuning

https://www.ibm.com/support/knowledgecenter/SSAW57
_8.5.5/com.ibm.websphere.nd.doc/ae/tprf_tunejvm_v6
1.html
AIX

http://www.redbooks.ibm.com/redbooks/pdfs/sg246478.
pdf

What should I Do to Improve Performance?

WebSphere Application Server

Java:

http://www.ibm.com/developerworks/websphere/librar
y/techarticles/0602_lurie/0602_lurie.html
http://wasdynacache.blogspot.com/2011/05/webspher
e-application-server-top-10.html
http://www.oracle.com/technetwork/java/tuning-139
912.html

MQ:

http://www.ibm.com/developerworks/websphere/libra
ry/techarticles/0807_hsieh/0807_hsieh.html

Troubleshooting

Structured thinking, repeatable use case


Narrow it down (find a pattern): Observation
and test cases
Fast, Nothing Changed, Now Slow
Best Practices and Logs: Often Inefficient by
Themselves

Serviceability

Baseline of What can be Realistically Expected

Common Items

Application: Number of Threads

Database: Buffer pools, Runtime Statistics

Probabilistic Use Cases: Big Buckets

WebSphere Properties
1) MDM JVM Heap size
2) ORB services maximum cache connections
3) ORB services thread pools
4) Default server thread pools
5) Server EJB cache size
6) Data Source JDBC connection pool
7) Data Source statement cache size
8) Transaction timeout

Questions...

You might also like