Performance Tuning for Content Manager

380 Performance Tuning for Content Manager

not debugging errors, set the Library Server stored procedure log trace level to

zero, which is the default setting.

It is the same for Resource Manager: Set the trace level to DEBUG (trace full)

only if you want to debug the issue in the Resource Manager server.

Often there is a requirement to generate logs and traces on a production system

at the failure point. IBM DB2 Content Manager Version 8.3 has a new feature that

generates tracing for a single user. In most cases, single-user tracing is enough to

get the details of the problem. Single-user tracing is enabled through the system

administration user interface. When it is enabled, other users get default ERROR

logging. It goes into effect at the next logon of the Content Manager user ID.

12.8.3 Applying fix packs

Performance improvements are often implemented in Content Manager software

fix packs. Check the fix pack readme document to find out whether it resolves

your performance issue.

For example, in one scenario, a Content Manager client application search on

“Documents and Folders” took much longer to return the search results list than a

search directly on either the “Documents Only” or “Folders Only” level. We later

found that a SQL statement required additional DB2 indexes, so we added three

additional database indexes on the Library Server tables in the new fix pack.

12.8.4 Information Integrator for Content API programming

Here is a technical note from the IBM Content Manager support Web site about

the general rule of Information Integrator for Content API programming. This site

has a lot of similar technical notes, so we encourage you to visit the site and view

them for more information.

Retrieving items that have links

You can streamline the process of retrieving items that have links by

remembering the following points:

 Do not retrieve links until necessary.

 Retrieve only the necessary direction of links.

Link retrieval spends time creating DKDDO objects for each retrieved link. A few

things can be done to avoid all or some of this processing time.

Do not retrieve links until necessary

In some cases, a link might not be needed at all or not needed until later when it

could be retrieved individually. Avoid retrieving links for a collection of items if