Hi... with db2exfmt it is sometimes a bit difficult to see exactly what
the subsections are... Many times analyzing queries, I will also generate
a dynamic explain plan using dynexpln... this clearly shows the
subsections and the numbers should match what is in the application
Rows read are all the rows that get scanned, not necessarily the ones that
are returned to the application... it is all the rows that had to be read
to obtain the result set. So if the rows read is very high, perhaps an
index could help, if tablescans are being done.
Rows written do not include table queue activity... it does include the
any table rows that are inserted, updated or deleted, including temporary
> I have had no luck mapping the subsections in an application snapshot
> with the query's explain plan output? The number of sections are the
> same, but the work being done seems to differ. For example, an
> explain plan may show a subsection 2 being broadcast to 16 partitions
> (EEE) and subsection 3 going to 1 partition. But the applicaition
> snapshot may show subsection 3 on 16 partitions and subsection 2 on 1.
> Also...any clarification on 'rows read' and 'rows written' would help.
> Does 'rows read include every row read on every table, whether it is
> part of the result set or not? Does 'rows written' include broadcast
> table queues in a EEE environment or is it intra-partition worktables
> and sorts only?
DB2 UDB Consulting Services
IBM Toronto Lab