whm@arizona.UUCP (Bill Mitchell) (07/03/84)
The swap spaces for Eagles on 4.2 are sized at 32M, a doubling of
the 4.1 16M size. The documentation presents no other rationale
for this increase other than the size of the disks (>400M). We have
two Eagles on our 780 (swapping on both), and we've never run out of
swap space. I've been considering using the 16M swap spaces ala 4.1
since they have proven to be adequate with 4.1 and because I'm not
ready to give up 32M for no good reason. However, I'm wondering if
the additional swap space would improve performace or if the extra
would just be wasted since the 2x16M we have now is adequate. Could
anyone offer advice on this point?
Bill Mitchell
whm.arizona@csnet-relay
{noao,mcnc,utah-cs}!arizona!whmobrien@CSNET-SH.ARPA (07/07/84)
From: Mike O'Brien <obrien@CSNET-SH.ARPA> I believe the increased size of the swap space merely reflects a growth in the size of the "average" program run on a BSD system. If you didn't run out of swap (or even come close) under 4.1BSD, then you can safely shrink the size of the swap partition under 4.2BSD. In fact when I did a 4.2 conversion recently I redid many of the partition sizes to reflect the fact that our partition sizes under 4.1 didn't resemble the 4.2 partitions in the slightest. Of course, you should use "vmstat" or something similar to determine how much swap you're actually using before you give any away.
mike@RICE.ARPA (07/09/84)
From: Mike Caplinger <mike@RICE.ARPA> So far as I can tell, the utilization of swap space has gotten poorer under 4.2. We frequently run out running the same job mix that worked fine under 4.1. Try a "pstat -s" and look at the "wasted" parameter; it gets large. I've heard this is because 4.2 now uses a fast but space-wasteful swap space allocator that generates lots of fragmentation. I don't know why they didn't leave it alone. Does anybody want to confess to doing this, or knowing what was done? - Mike