Q and A

Asked and Answered

We're starting a new branch project built on Unix and WebSphere Application Server with WebSphere Business Components Composer (WSBCC). The project will produce 3,000,000 logs daily with 1,000 concurrent users, and the staff wants to put the logs on DB2 for OS/390. WSBCC uses JDBC and dynamic SQL. Is the DB2 Distributed Data Facility (DDF) designed for such a heavy process? What possible problems should we be aware of?

Robert Catterall responds:

I believe DDF could in fact handle such a workload. You'll want to be on at least DB2 for OS/390 version 6 (v.7 would be the best) and a fairly current release of the OS/390 or z/OS operating system. (I'd suggest at least OS/390 version 1 release 10, if not z/OS). Some considerations:

• Use dynamic statement caching to minimize the CPU cost of dynamic statement preparation. This may consume a lot of virtual storage; so consider putting the part of the EDM pool used for the dynamic statement cache in a data space address space (rather than in the DBM1 address space). Longer term, consider moving from JDBC to SQLJ, so that you can take advantage of static SQL (may not be an option if the JDBC is in vendor-supplied code).
• Consider moving to an architecture that has most of the table-touching SQL (SELECT, INSERT, UPDATE, DELETE, OPEN CURSOR) in DB2 stored procedures. The two advantages here are lower network overhead costs and the opportunity to use static SQL. Consider using SQL stored procedures for this purpose - they're pretty easy to develop and deploy. The client programs can use JDBC to call the stored procedures and to manipulate output parameters and result set values returned by the stored procedures.
• Set up your mainframe DB2 subsystem for inactive DDF thread support. This will enable you to efficiently support a very large number of downstream connections.
• Take advantage of the connection pooling and concentration capabilities of DB2 Connect version 8. This will help you get the most efficient use of your connections to the mainframe DB2 subsystem.

You can certainly support a workload volume in the hundreds of transactions per second using DDF. It's mature technology at this point, with many large workloads in production at numerous mainframe DB2 sites.

See a complete archive of reader/author Q&As.