X-no-Archive:yes
Changing the PCs from 486/Win95 to Pentium/NT, in and of itself will not
make a difference to the HP 3000, or any other server.  What can happen
though is an increase in work from the user, occasioning an increase of
activity on the client and the server.  Windows NT is a higher performance
OS than Win 95 and the Pentium II boxes perform a lot better than the older
Pentiums and the 486s.

So, with the little bit of information in the enclosed message, I would
venture to say the users are getting a lot more work done on their PCs and
this is causing more work for the HP 3000 :->.

Then again maybe something is configured wrong somewhere.

Kind regards,

Denys. . .

Denys Beauchemin
HICOMP America, Inc.
(800) 323-8863  (281) 288-7438         Fax: (281) 355-6879
denys at hicomp.com                             www.hicomp.com


-----Original Message-----
From:   Trudeau, James L [SMTP:[log in to unmask]]
Sent:   Tuesday, 16 June, 1998 10:37 AM
To:     [log in to unmask]
Subject:        Restore degradation - resolved

Howdy,

Ah, tis a bright and shiny day, my restore probelm went away.
Many thanks to those who contributed.  What follows is
either a FAQ or a lesson to others.

One of the first questions asked was about disc space.  We
have 22GB mounted with 5.5GB free at any given time.  The
restoreset giving us the problem was 5.2 GB consisting of
1 database of 195 datasets.  About 25% of the files are under
5 MB the largest is about 2GB with the rest filling in the gaps
at 100 to 500 MB (all that is from memory so don't add it up
please).  When the system began shutting down VT sessions
it did not appear that disc was a problem.  So.........

Last night we doubled memory from 192 MB, where it's been
for three years to 384 MB.  This morning I ran the "restore
from hell" and took a look at memory usage during the
especially critical last 15 minutes.  The results:

cpu was unchanged.......big suprise huh?
transient dropped from 30/sec to 6/sec
file object dropped from 5.2/sec to 1.5/sec
NM stack dropped from 10.6/sec to 2.1/sec
clock rate dropped from 78 to 34
page faluts from 45/sec to 10.5

No one got bumped off the system and the "irritating" response time,
although still noticible, is now negligible.  Looks like VTSERVER was
eating our lunch.

The answer to "how did this happen so suddenly" also became
apparent........sort of.  I knew we had been "pushing the knee"
to a degree for some time.  What happend was that to me the
number of users logged on appeard to be within the usual range, but
my cohorts on the desktop side had gone out and upgraded a bunch
(15-20 PC's) from whatever with WIN95 to Pentium somethings with
NT.  Now why the difference between WIN95 and NT should show
up on the HP 3000 by crashing VT sessions?  I will leave that
explanation to Denys, et al as it is beyond me.

I was also asked why we have such a "big box" for so few OLTP
users.  We don't need it to support the online folks, we could still
be running our old 935 for that, we need it for nightly MRP and
supporting batch runs.

Anyhow, i guess the moral of the story is:  Keep an eye on yer
memory size/usage.  Just like everyone keeps telling you to.

Again thanks to all.  Now send rain please.

Jim (cookin') Trudeau
CSC
Harlingen, Tx