 |
The Question is:
Is the scheduler, e.g. the usage of quantum/iota significantly changed in
AXP/VMS V7.2 compared to VAX/VMS 6.2....?
It can be demonstrated, that the dynamics of OpenVMS are completely different
from
the "old times". If there is only ONE compute bound job (prio=4) without any
i/o or
paging on the system (DS10), then the behavior of the system is extreme
sluggish, e.g. mc sysgen show quantum does take 40 seconds in a second
decterm? (quantum is 10, iota=1).
I did encounter the same kind of problems on axp 4100 multiple cpu mashines,
but never had a chance, to track the user (stock traders) complaints in any
way, so we had been having this kind of non-traceble problem for more than 2
years, most of the time,
the extra cpu's on the 4100's did circumvent the problem.
The Answer is :
It can be demonstrated that OpenVMS has changed significantly from
"old times" in a variety of ways. Old rules-of-thumb and older
tuning techniques may or may not be applicable to newer OpenVMS
releases and/or to particular OpenVMS VAX or OpenVMS Alpha systems.
As for one discussion of a change related to the processing around
QUANTUM, please see topic 4341 here in Ask The Wizard.
The value of IOTA defaults to 2, and changes to it generally also
involve a (recommended) discussion with Compaq Customer Services.
Without significantly more information on the system configuration
(hardware, software, settings, etc), the OpenVMS Wizard is not in a
position to assist with tuning and site- and configuration-specific
issues. Please contact the Compaq Customer Support Center directly
for assistance in tuning the system, and for assistance meeting the
particular system responsiveness requirements.
|
