Re: CPU Increase in 16.2

Discussion created by ca.portal.admin on Oct 1, 2005
Chris and Listers -

Chris - Thank You again for the information you have shared.

I do not really have a conclusion on this.

I do have several things to say:

C.A. and I (us) cannot come to agreement that there has been an increase
in CPU usage. We FTP'd PMIM stat records and they have run their own
reports and summarized their findings.=20
They have concluded NO increase.

I do not agree with the results because the conclusions were based on
average CPU utilization per task comparing one day of 14.1 to another on
16.0 and repeating the comparisons for several other selected days.
Although they say that the days were selected based on similarities in
number of task executions, DB calls, I/O, etc., I could not help but
notice that the days they selected had high average CPU for the 14.1
data and low average CPU for the 16.0 data. I could not help but get
the impression that the days they selected were intended to be days that
would produce the desired outcome.

I would prefer to wait until we have a month's worth of 16.0 data (still
another 1-1/2 weeks yet) and then compare a month of 16.0 averages to a
month of 14.1 averages. The business is not seasonal, so I would expect
that a comparison of a full month's worth of data would produce
statistically more accurate results.

C.A. has said that British Telecom did a benchmark study that showed NO
Perhaps anybody from there on this list might want to chime in.

My intent is not to alarm everybody.
My intent is to seek help as to why.

I haven't dismissed that something defined here is wrong or requires

Also, I want to be clear that if the increase is because 16.0 is
functionally more robust that 14.1, we can live with that.
If it is because something is defined wrong here, that's fine, too. =20
If I know what's wrong, I can remedy the condition.=20

Where our difficulty lies is that we cannot come to agreement on the
statistics at-hand and to-date.
Two different groups are looking at the same data and drawing two
entirely different conclusions. =20

Jon Gocher

----- Original Message -----=20
From: Wood, Chris=20
To: Jon Gocher=20
Sent: Friday, September 30, 2005 1:03 PM
Subject: RE: CPU Increase in 16.2

Any news yet Jon?




IDMS Public Discussion Forum


FW: CPU Increase in 16.0
"Jon ,

I can confirm that we found zero overall impact on CPU between 15.0 and
16.0 on our test benchmarks ( using TPNS) and in two pilot production
systems .

BT plc