DSXchange

hi,

If we are asked to design a job given with a scenario how to decide whether we should go for server job or parallel job?

What are the essentials in considering server or parallel job with respect to both unix or windows platform.

Thanks,
sravanthi

The number of records, the size of a single record, what operation needs to be performed on the data, the transformations, the time window for completion; just to name a few.

If you only have server edition licensed the decision is a no-brainer. Likewise if you've chosen only to license only enterprise edition.

Small jobs - server.
Large jobs - parallel (probably).
Medium jobs - some other factor needed to make the decision.

Nothing quantifiable!

There's a lot of parallelism you can build into server jobs: multi-instance, row buffering, parallel streams, link partitioning, etc., just to muddy the waters. I have one long-standing client processing huge volumes of data with multi-instance server jobs, and quite happy with the performance.

The interface between server jobs and IBM Information Server - how efficiently the conversion occurs at the boundary - might also be a consideration for version 8.0 and beyond.

Ray,

The client whom you say of processing huge volumes using Server jobs, what is the approximate volume. Are they processing up to 10 or 15 million?

Thanks.

vnspn wrote: The client whom you say of processing huge volumes using Server jobs, what is the approximate volume. Are they processing up to 10 or 15 million?

No issues with Server jobs handling such volumes. The approach/requirement for your processing can give you different results.