Oppolzer - Informatik / Blog

There are indeed plans to migrate to RRS interface.
For the moment there is no problem, because this region is read-only.
The DB update regions are IMS and DB2, without MQ.

In the meantime I found the reason for the high CPU consumption.

The interface DSNALI chases the CDE chain to find out, if the modules
DSNACAB and DSNACAF are already loaded. This is done on every call
to DSNHLI (that is, DSNALI). This may be ok, if there is one batch program
and a very short CDE list. But our CDE list has thousands of modules.
This is the problem.

Do you think that we will get IBM to fix this and give us an improved version
of DSNALI, or will I have to tune it myself? I guess, little hope for a global
solution. I'll try to fix it without changing DSNALI, but I'm still looking for
a clever idea.

BTW: 4 percent CPU means hundreds of thousands of Euros per year for us.

Kind regards

Bernd



M.K. schrieb:
>
> Do you get two phase commit between the resources involved (DB2, IMS,
> MQ) using DSNALI ? IIRC the answer is no in which case I'd be more
> concerned about data inegrity than about 4.9% extra cpu. Or this is
> some special case where you don't care about proper commits/rollbacks ?
>
> Khalid