[sqr-users] TCP v. BEQ

["George Jansen" <GJANSEN@aflcio.org>]

Yelena,

Brian Johnson, a minicomputer performance guy, used to say that there
were two questions:

1. Does it make a difference?
2. Does it matter?

I'm suspecting that the answers are Yes, and No. I would offer you
three suggestions--I suspect that the chance of timely success is
inverse to cost, but in order of cost ($$ and effort).

1. Bring in an SQR consultant, Brio or 3rd-party. Not cheap, but likely
to be effective, and probably worth it if you are missing your windows.

2. Copy tooslow.sqr to tooslow2.sqr and go through the latter,
commenting out EVERYTHING that is not hitting the database
(select/update/insert) or calling another procedure that hits the
database. How fast does that run? If it runs a lot faster, and I bet it
will, starting uncommenting till you hit the trouble spots.

3. Do a quick-and-dirty profile of your SQR, timing all procedures
EXCEPT where they are in a begin-select or begin-sql paragraph. Maybe
the Ontko gang have something on their site to do that. You can do a
blunt-force version of this with a fairly short script--feed it an SQR,
get a profiling SQR. What does your profiled SQR show you?

I must confess I don't understand your having so much time in "SQL*Net
message TO client". I'm able to manipulate "SQL*Net more data to client"
by increasing my -B value, but the other doesn't budge from a very low
value.

What versions of PeopleSoft, SQR, and Oracle?


George Jansen


_______________________________________________
sqr-users mailing list
sqr-users@sqrug.org
http://www.sqrug.org/mailman/listinfo/sqr-users