Performance Tuning for Content Manager

348 Performance Tuning for Content Manager

12.1 Troubleshooting basics

The key to successfully and efficiently resolving performance problems in a

Content Manager system is to have an organized, disciplined process that

focuses on identifying and relieving the system’s performance bottlenecks.

Without an organized process focused on identifying bottlenecks, it is very easy

to spend a lot of time and energy tweaking various parameters, ending up with a

system in a worse state than before with no good record of what was changed

and why. To be successful, you must collect the right data and find the real

bottleneck. You will then be able to address that bottleneck, tune the system, and

improve the performance of your Content Manager system.

Before beginning any full performance troubleshooting process, first ensure that

the database statistics for the Library Server and Resource Manager databases

are up to date. Content Manager is a database application. It depends heavily on

accurate database statistics for optimizing all operations. Updating database

statistics on the Library Server and Resource Manager databases is quick and

easy to do, and often turns out to be the only fix needed.

Performance problems are often caused by changes in the system. The changes

that might affect system performance include:

 Hardware changes

Example: Adding new disks or changing hardware configurations

 Operating system changes

Example: Installing PTFs or changing parameters

 Software changes

Example: Adding fix packs for DB2 and WebSphere, or configuration changes

 Application changes

Example: Installing or upgrading new versions and fixes, configuration

changes, installing new modules, or adding new users or new content

 Performance tuning

Example: Tuning changes in operating system, network, DB2, WebSphere,

TSM, or the application

 Any changes

Is it a performance problem?

When a user complains about a problem or you experience a problem firsthand,

you must first determine the nature of the problem: whether it is a functional

problem or a performance problem. When your hardware, network, or application

does not behave correctly, it is a functional problem (for example, if your Content

Chapter 12. Troubleshooting performance problems 349

Manager application has a memory leak). Sometimes, functional problems lead

to performance problems (an application memory leak eventually slowing down

the overall system performance). In such cases, it is important to determine the

root cause of the problem and fix it rather than focus on tuning your system.

Using the example of application memory leak, you need to address the problem

to the development team and ask them to eliminate the memory leak.

Asking questions

Before jumping into collecting data and analyzing data, you should get as much

detailed information as possible about the problem experienced by users.

Questions to ask users or to research on your own include:

 Is this an across-the-board problem, or does it affect only certain operations

or users?

 What operations are slow — the time it takes to search for a document or to

display a document, for example?

 Can the problem be reproduced by exercising a certain task or sequence of

events?

 Is the slow performance intermittent (does it disappear from time to time)?

Does it occur at a certain time of day or in relation to certain tasks?

 When did the problem start occurring? Was the situation the same when the

system was first installed or went into production? (Is this a regression or

were the performance objectives never met?) Did anything change on the

system before the problem occurred (such as adding more users or loading

additional legacy content to the system)?

 Is this a single-user performance problem or multi-user scalability problem?

 If remote users are complaining about document retrieval, do local users

performing the same document retrieval experience the same slowness

(network versus server issue)?

 If this is network related, how are the network segments configured (including

bandwidth such as 10 Mbps or 9600 baud)? Are there any routers between

the client and server? Are the objects cached?

 What other applications are running on the system, and are those

applications involved in the performance issue?

 What is the impact of the performance problem on the users?

After understanding the nature of the problem, we need to find the location of the

bottleneck or what area causes the problem. Several actions can be taken:

 Use monitoring tools and programs to check system performance:

– Obtain workload metrics to compare with the baseline.