Friday 4 March 2016

I need to do it, I need to document the undocumented setting–faster JDE on oracle

I’ve been sworn to secrecy regarding this setting, but it’s gotta come out.  Are you having UBE’s with unpredictable performance, fast one minute and then slow the next.  Fast one user and slow the next?  You need to get into a bit of this action:

alter system set “_optim_peek_user_binds”=FALSE scope=both;

This is for oracle, you might be able to tell.

I’ve used this at a number of clients with complete success.  I’ve also worked on an OMCS managed site, where this is a recommendation from them for JDE – note that you might not find a MOS reference.

There are a number of references on the web, but it seems to make the QEP (query execution plan) more predictable for JDE.

Let me know if it kicked any goals for you!  How?  Search my blog for F986114 and there are some good statistical queries that you could run to answer this question.

1 comment:

Bill said...

I'm trying this out at a couple of clients now, one on 11g and AIX/Power8 and another on Supercluster and 12c. I imagine this improves overall performance, but do you think that when you look at individual UBE or interactive performance you'd see most get better and a few get worse?