On 97-03-06 12:38:58 EST [log in to unmask] (Tom Haws) wrote

> Hi list,
>
>Our users are concerned with on-going performance monitoring on our 987
>running 5.0.  They would like to see something set up that gives some sort
>of a performance metric, that could be monitored as barometer of system
>performance.  We already have Laser/RX, but they feel that is too abstract,
>and may check low-level system performance, but does not give a good
>indication of an actual user's response as they interact with the machine.
>They would rather see something that checks a process logging in, doing
>something user-ish (a large file copy, sorting a dedicated database etc.),
>and reporting a number.  We could calibrate the "barometer" based on history
>and experience, so that when the number this magical process reports hits a
>certain level, the machine gets attention from Systems staff, hopefully
>before our HelpDesk gets flooded with "the system is slow" calls....
>
>Does anyone know of a freeware utility or have a custom jobstream that we
>could get to wake up every hour and do a quick check of the machine?  We
>have HP IT Operations Centre agents loaded, so it could report to the 987
>system console and thence to our ITO console, which is always manned.
>
>If there are commercial products or consultants out there able or willing to
>do this, we would also be interested, but this task was not foreseen in this
>year's budget, so may be hard to justify.  (Can you say PLAN THE BUDGET?)
>
>I will summarize.
>-Tom Haws

A former employer with a simlar problem integrated the collection of the
elapsed time between certain transactions and collection of database
statistics into their order entry and processing applications.  They were
taking and shipping over 50,000 orders per day across 22 warehouses running
on 7 HP3000s.  The collected stats were summarized and graphs published every
day.  The user words "response is bad" always corresponded to the charts.
 The chart included response time, order lines, lines printed delay, etc. on
1/2 hour increments.  I'm not suggesting that this is better than the system
measurement tools, but rather that this is a good way to delelop a service
level agreement with the users that can be measured and audited in their
terms.  Internally to systems, they still used all the tools like Laser/Rx
and Glancexl, but used separate 'systems' charts that usually tracked and
helped explain what the users would see and feel.

Hope this helps.


Chuck Duncan
[log in to unmask]