I'm running the NetBSD-based Xkernel on my Sun 3/60.  It is served and
serves a SPARCClassic running Solaris 2.5.  I just noticed that it
seems to be creating a significant load on the SPARCClassic.  Is this
normal?

When I have the 3/60 running, but not doing anything (for example,
waiting for a login (XDM, or whatever the CDE equivallent is), there
is a significant load.  When I type L1-A on the 3/60, taking it out of
contention, the load is reduced.

Here are some log entries from Solaris' perfmeter.  Can you guess when
I typed L1-A?

05/29/97 23:43:03 localhost 	cpu=12.00	pkts=289.00	intr=303.00	cntxt=312.00
05/29/97 23:43:05 localhost 	cpu=14.00	pkts=288.00	intr=288.00	cntxt=309.00
05/29/97 23:43:07 localhost 	cpu=12.00	pkts=290.00	intr=303.00	cntxt=311.00
05/29/97 23:43:09 localhost 	cpu=18.00	pkts=288.00	intr=300.00	cntxt=309.00
05/29/97 23:43:11 localhost 	cpu=13.00	pkts=289.00	intr=290.00	cntxt=308.00
05/29/97 23:43:13 localhost 	cpu=15.00	pkts=290.00	intr=301.00	cntxt=310.00
05/29/97 23:43:15 localhost 	cpu=16.00	pkts=288.00	intr=301.00	cntxt=309.00
05/29/97 23:43:17 localhost 	cpu=17.00	pkts=290.00	intr=291.00	cntxt=309.00
05/29/97 23:43:19 localhost 	cpu=19.00	pkts=286.00	intr=301.00	cntxt=309.00
05/29/97 23:43:21 localhost 	cpu=14.00	pkts=289.00	intr=301.00	cntxt=308.00
05/29/97 23:43:23 localhost 	cpu=18.00	pkts=290.00	intr=292.00	cntxt=314.00
05/29/97 23:43:25 localhost 	cpu=2.00	pkts=16.00	intr=28.00	cntxt=37.00
05/29/97 23:43:27 localhost 	cpu=0.00	pkts=4.00	intr=17.00	cntxt=25.00
05/29/97 23:43:29 localhost 	cpu=1.00	pkts=4.00	intr=6.00	cntxt=27.00
05/29/97 23:43:31 localhost 	cpu=0.00	pkts=3.00	intr=16.00	cntxt=23.00
05/29/97 23:43:34 localhost 	cpu=1.00	pkts=3.00	intr=16.00	cntxt=24.00
05/29/97 23:43:36 localhost 	cpu=3.00	pkts=4.00	intr=19.00	cntxt=29.00
05/29/97 23:43:38 localhost 	cpu=0.00	pkts=2.00	intr=15.00	cntxt=25.00
05/29/97 23:43:40 localhost 	cpu=1.00	pkts=2.00	intr=15.00	cntxt=26.00
05/29/97 23:43:42 localhost 	cpu=0.00	pkts=2.00	intr=3.00	cntxt=24.00
05/29/97 23:43:44 localhost 	cpu=0.00	pkts=2.00	intr=16.00	cntxt=26.00

cpu     Percent of CPU being utilized.
pkts    Ethernet packets per second.
intr    Number of device interrupts per second.
cntxt   Number of context switches per second.

[Other things metered by perfmeter were not interesting.]

Is this reasonable?  Normal?  To be expected?  Easy to improve?

Hugh Redelmeier
hugh@mimosa.com  voice: +1 416 482-8253