RE: Verifier Performance

Hi Reeve,


> IBM has told us the communications manager is going to be improved; that's
> going to relate to how QZRCSRVS works for member locking and Verify
> performance.  My guess is that we're not going to be seeing anything until
> the next OS/400 release unless the QZRCSRVS replacement is PTF'ed.  I
> haven't heard of anything new in V5R2, so I'm assuming an
> improvement is in
> the future.

It seems to me that the biggest problem is not locking, but the sheer number
of QZRCSRVS jobs being cycled. As you're aware, starting a new job is a
resource intensive operation in OS/400. When I verify, I can see dozens of
these jobs being used, since each is only good for a single command. The job
runs for a fraction of a second before ending. It takes much longer to start
and end the job than it does to peform the work it was asked to do.

In fact, I'm truly wondering if there isn't something wrong with my
configuration, because I'm having trouble believing that Code is supposed to
use QZRCSRVS as much as it's doing on my machine. That's got to be one of
the worst ways to do client server on an AS/400, and I'm quite certain that
the Code team knows a lot more about writing a client server app than I do.


> This may not be much consolation, but the ability to compile from the IFS
> may be IBM's oblique solution to Verify performance; I haven't
> experimented
> with it because I'm still on V5R1.

I hope that's not the answer. I have so much code in PF-SRC members that it
would be a monstrous task to convert everything to stream files. Besides,
what happens to the attributes (source type, member description) if you go
to a stream file? Let me guess... we'd lose them in the name of "progress".
:)


John Taylor