On Fri, 2 Aug 1996 03:43:58 +0200 (MET DST), Herbert Rosmanith
<herp@wildsau.idv.uni-linz.ac.at> said:
>> > nb: obviously, netscape crashed the day before (1376 div 60 = 22), and
>> > was since then consuming nearly all of the CPU% ... maybe the scheduling
>> > algorithm could be adapted to give lower priority to processes
>> > with high cpu-utilization
>>
>> It does.
> have you ever tried working on a machine that is running a
> tape-backup ? (scsi,aha2940, for example) ...
Yes, frequently. I hardly notice the load.
That's probably because I'm backing up from EIDE with triton DMA onto
DAT. DMA all the way, so there's little cpu load. I've also got 32MB
on this box, so I'm doing very little swapping.
However, you often would expect to see bad performance while backing
up. It's got very little to do with the scheduling algorithm, and
everything to do with the fact that if the disk is occupied for
backup, then any paging or swapping from the same disk is going to be
*painfully* slow. If you are running IDE, then access to the other
drive of the master/slave pair being backed up is also going to suffer
greatly.
Cheers,
Stephen.
-- Stephen Tweedie <sct@dcs.ed.ac.uk> Department of Computer Science, Edinburgh University, Scotland.