[MLton] Paging under RedHat 9.0 although largely fitting in core (resend)
Alain Deutsch
deutsch@PolySpace.com
Tue, 09 Mar 2004 17:09:37 +0100
We have noticed the following performance problem under RedHat 9.0:
there are situations where paging seemingly occurs under RedHat 9.0
although the MLton-generated executable completely resides in core. In
the following instance, I have set ramslop to a low 0.75, and despite
this and the fact that memory is largely unoccupied, the % of CPU use is
closer to 50% than to 90%, as I would expect.
Enclosed is a top snapshot: we can see that:
1. the MLton-generated executable, iabc-c2if.*, only uses 44% of the CPU
2. its size, 762Mb, is much less than the total RAM size (1Gb)
3. the core is largely free besides the MLton-generated executable
4. kswapd is active and eats up 25% of CPU time
I was wondering:
1. if this is due to RedHat 9.0
2. if so, whether the use of madvise() calls from the gc could ease
up the situation (perhaps by using WONTNEED on old from spaces ?)
3. if there may be RedHat 9.0 kernel and/or swapd setting to avoid
this behavior.
Thanks.
10:43:30 up 18 days, 20:42, 2 users, load average: 2.27, 2.14, 1.89
53 processes: 49 sleeping, 4 running, 0 zombie, 0 stopped
CPU states: 23.2% user 75.6% system 0.0% nice 0.0% iowait 1.0% idle
Mem: 1030404k av, 1020248k used, 10156k free, 0k shrd, 5288k
buff
850308k actv, 50612k in_d, 20932k in_c
Swap: 3849912k av, 694424k used, 3155488k free 168992k
cached
PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME CPU COMMAND
881 deutsch 18 0 762M 751M 1404 R 44.8 74.6 15:18 0
iabc-c2if.cx86-
5 root 16 0 0 0 0 RW 25.9 0.0 52:12 0 kswapd
1019 deutsch 16 0 1252 1252 928 R 6.4 0.1 0:06 0 top
15 root 16 0 0 0 0 DW 6.3 0.0 4:31 0 kjournald
32173 deutsch 15 0 528 232 124 S 1.4 0.0 0:02 0 sshd
966 deutsch 15 0 1880 1712 1484 S 1.0 0.1 0:00 0 sshd
1027 deutsch 16 0 560 560 476 S 0.5 0.0 0:00 0 tail
2 root 15 0 0 0 0 RW 0.1 0.0 0:11 0 keventd
1 root 15 0 88 60 40 S 0.0 0.0 0:06 0 init
3 root 15 0 0 0 0 SW 0.0 0.0 0:01 0 kapmd
4 root 34 19 0 0 0 SWN 0.0 0.0 0:00 0
ksoftirqd_CPU0
9 root 15 0 0 0 0 SW 0.0 0.0 0:00 0 bdflush
6 root 15 0 0 0 0 SW 0.0 0.0 0:03 0 kscand/DMA
7 root 16 0 0 0 0 SW 0.0 0.0 55:11 0
kscand/Normal
8 root 15 0 0 0 0 SW 0.0 0.0 5:32 0
kscand/HighMem
10 root 15 0 0 0 0 SW 0.0 0.0 0:02 0 kupdated
11 root 25 0 0 0 0 SW 0.0 0.0 0:00 0
mdrecoveryd
72 root 25 0 0 0 0 SW 0.0 0.0 0:00 0 khubd
3690 root 15 0 212 164 112 S 0.0 0.0 0:03 0 syslogd
3695 root 15 0 52 4 0 S 0.0 0.0 0:00 0 klogd
3721 daemon 15 0 272 248 184 S 0.0 0.0 0:36 0 portmap
3799 root 15 0 0 0 0 SW 0.0 0.0 0:17 0 rpciod
3800 root 25 0 0 0 0 SW 0.0 0.0 0:00 0 lockd
3812 root 24 0 52 4 0 S 0.0 0.0 0:00 0 apmd
3828 root 25 0 304 260 188 S 0.0 0.0 0:02 0 ypbind
3910 root 15 0 280 216 196 S 0.0 0.0 0:00 0 automount
3912 root 15 0 112 48 28 S 0.0 0.0 0:00 0 automount
3928 root 15 0 352 112 108 S 0.0 0.0 0:14 0 sshd
3942 root 24 0 128 4 0 S 0.0 0.0 0:00 0 xinetd
3981 root 16 0 13064 4 0 S 0.0 0.0 0:00 0 spamd
3990 root 15 0 48 4 0 S 0.0 0.0 0:00 0 gpm
4000 bin 15 0 460 40 12 S 0.0 0.0 0:00 0
cannaserver
4011 root 15 0 176 132 88 S 0.0 0.0 0:03 0 crond
4022 root 15 0 1748 260 224 S 0.0 0.0 0:07 0 cupsd
4067 root 25 0 3356 4 0 S 0.0 0.0 0:00 0 jserver
4144 xfs 15 0 4048 44 24 S 0.0 0.0 0:00 0 xfs
4179 root 23 0 48 4 0 S 0.0 0.0 0:00 0 mingetty
4180 root 23 0 44 4 0 S 0.0 0.0 0:00 0 mingetty
4181 root 23 0 44 4 0 S 0.0 0.0 0:00 0 mingetty
4182 root 23 0 44 4 0 S 0.0 0.0 0:00 0 mingetty
4183 root 20 0 44 4 0 S 0.0 0.0 0:00 0 mingetty
4184 root 20 0 44 4 0 S 0.0 0.0 0:00 0 mingetty
32171 root 21 0 364 0 0 SW 0.0 0.0 0:00 0 sshd
32174 deutsch 15 0 788 640 432 S 0.0 0.0 0:00 0 bash
447 deutsch 15 0 60 8 0 S 0.0 0.0 0:00 0
rte-kernel.x86-
474 deutsch 15 0 260 8 0 S 0.0 0.0 0:00 0
rte-kernel2
548 deutsch 15 0 76 60 28 S 0.0 0.0 0:00 0 tee
783 deutsch 15 0 132 4 0 S 0.0 0.0 0:00 0 rte-suif
877 deutsch 15 0 44 4 0 S 0.0 0.0 0:00 0
rte-time.x86-li
964 root 20 0 1640 1432 1276 S 0.0 0.1 0:00 0 sshd
967 deutsch 16 0 1516 1472 1180 S 0.0 0.1 0:00 0 bash
--
Alain Deutsch, CTO
PolySpace Technologies
mailto:deutsch@POLYSPACE.COM http://www.polyspace.com