home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.os.vms
- Path: sparky!uunet!spool.mu.edu!uwm.edu!ux1.cso.uiuc.edu!news.cso.uiuc.edu!jsue
- From: jsue@ncsa.uiuc.edu (Jeffrey L. Sue)
- Subject: Re: TUNING - QUANTUM SYSGEM parameter
- References: <27A0322A_00174E60.00963CCA9D4FA020$626_2@UK.AC.KCL.PH.IPG>
- Message-ID: <1992Nov19.134754.25572@ncsa.uiuc.edu>
- Originator: jsue@troon.ncsa.uiuc.edu
- Sender: usenet@news.cso.uiuc.edu (Net Noise owner)
- Organization: The Dow Chemical Company
- Date: Thu, 19 Nov 1992 13:47:54 GMT
- Lines: 57
-
- In article <27A0322A_00174E60.00963CCA9D4FA020$626_2@UK.AC.KCL.PH.IPG> SYSMGR@IPG.PH.KCL.AC.UK writes:
- >Slightly provocative reply:
- >
- >If QUANTUM=20 is OK on a microVAX-II (less than 1 VUP) (and it is!), then
- >logically QUANTUM=1 on anything rated at more than 20VUP should be OK,
- >maybe even optimal.
- >
-
- Maybe this information has already been posted. If so, I pologize.
-
- I have been to courses, and many DECUS sessions where Patrick O'Malley
- discusses tuning quantum. What should be kept in mind when reducing
- QUANTUM is that it may seemd to improve *trivial* performance, it reduces
- "throughput". This means that in the same amount of wall-clock time,
- you get less "total" work completed. This is a trade-off.
-
- Other factors to consider are: whenever a context switch takes place,
- the on-cpu instruction cache is invalidated. Thus the more often you
- context switch (smaller quantum), the more often the cache will become
- invalidated. Therefore your job will get less done because it must rebuild
- the cpu's instruction cache everytime it becomes the CUR processes, and it
- is doing this more often. (note: the CPU cache is a hardware cache, not
- one of the XQP (ACP_*) caches - which are for the file system.)
-
- I have played with quantum and found that on the smaller processors it
- made a big difference in the performance of "trivial" response, like for
- MAIL, DIRECTORY, DELETE, TYPE, etc.. However, on the faster processors
- today, rarely do these activities even use a quantum before they give up
- the cpu voluntarily for a wait-state (I/O completion, LEF, etc.). Also,
- in the cases where I lowered quantum, the longer-running operations tended
- to take more time - my guess is this happens because the scheduler must stop
- processing more often to determine "who's next" to run... even if there is
- only one job on the system.
-
- I eventually settled back to DEC's default of 20 for quantum. In the
- evenings I set quantum to 100 (1 full second) so that the long,
- number-crunching jobs can actually get more time in the cpu before the
- scheduler stops the processing to do a context switch. This can cause
- longer delays for interactive processing, and trivial response.
-
- Much of the decision on whether to reduce quantum depends on who/what
- you're trying to tune for. In my case I have a mix of trivial, and non-trivial
- processing going on at the same time. Thus I don't want to context-switch
- too often. I hesitate to say that, merely because today's processors are
- much faster than a uVAX II, you can lower quantum. One of the reasons we
- pay for the faster processors is to get more work completed. Throttling
- it all back to a uVAX II speed would miss this point. We could get the
- same affect by merely buy a lot more uVAX II's, distribute the load, and
- spend a lot less money than a VAX 6610.
-
- Hope this doesn't confuse the issue too much.
-
- Jeff
- --
- -----
- Jeff Sue
- - All opinions are mine - (and you can't have any, nya nya nya)
-