Subject: FYI -- "systat vmstat" from -current on 1.6.x.....
To: NetBSD/alpha Discussion List <port-alpha@NetBSD.ORG>
From: Greg A. Woods <woods@weird.com>
List: port-alpha
Date: 11/22/2004 16:34:53
Well I finally got fed up with the old "systat" brain-damage on my SMP
alphas because it wasn't really telling me anything terribly useful that
I couldn't see other ways and it was getting increasingly more difficult
to tune these systems properly as all the other tools are so much more
painstaking to use (and some features just aren't available either).
So, I ported the -current version back to the 1.6.x branch (which first
required porting libcurses back too, but that was trivial), and now I
must say I'm really quite impressed with the new "systat", especially
the new "systat vmstat" display. It's really nice to finally see the
interrupt rates on the alpha, and most amazing to finally see the full
system capacity represented properly for multi-CPU systems in the CPU
bargraph and CPU percentages!
I'm really looking forward to 2.1 (I never touch a .zero release! :-),
but my production environments really need a stable base and so I'm
really glad I took the time to do this "pullup" for myself!
If anyone would like the patches to do a pullup on the official
netbsd-1-6 branch, or for yourself, please just let me know!
This is with a build running and with NFS serving a build on another
machine and sundry other things like the emacs I'm typing to, named,
ntpd, etc. -- ccd0 is a striped array of sd0-sd3 and its where the
OBJDIR and DESTDIR are:
26 users Load 3.63 3.62 3.42 Mon Nov 22 16:19:29
Proc:r d s w Csw Trp Sys Int Sof Flt PAGING SWAPPING
2 3 28 337 2583 5990 3374 363 2452 in out in out
ops
44.8% Sy 24.2% Us 0.1% Ni 0.6% In 30.3% Id pages
| | | | | | | | | | |
======================>>>>>>>>>>>>> 28 forks
8 fkppw
memory totals (in kB) 4053 Interrupts 8 fksvm
real virtual free soft serial pwait
Active 963888 963888 225256 19 soft net relck
All 1306552 1306552 2273504 12 soft clock rlkok
1201 cpu0 clock noram
Namei Sys-cache Proc-cache 137 cpu0 device 229 ndcpy
Calls hits % hits % 181 cpu0 ipi 55 fltcp
5901 5877 100 6 0 microset ipi 1491 zfod
147 shootdown ipi 83 cow
Disks: seeks xfers bytes %busy 32 cpu0 imb ipi 32 fmin
ld0 2 synch fpu ipi 42 ftarg
ld1 discard fpu ipi itarg
ld2 1197 cpu1 clock 466 wired
ld3 487 cpu1 ipi pdfre
fd0 1 microset ipi pdscn
cd0 456 shootdown ipi
sd0 24 251K 3.5 29 cpu1 imb ipi
sd1 25 261K 2.4 synch fpu ipi
sd2 26 293K 3.3 1 discard fpu ipi
sd3 20 240K 2.2 42 kn300 irq 12
md0 104 kn300 irq 16
md1 kn300 irq 36
ccd0 92 1045K 5.0 5 kn300 irq 44
isa irq 4
and again a bit later with some activity on the system and sources disks:
26 users Load 4.00 3.84 3.56 Mon Nov 22 16:22:56
Proc:r d s w Csw Trp Sys Int Sof Flt PAGING SWAPPING
1 2 23 165 817 2468 2435 131 725 in out in out
ops
16.4% Sy 42.9% Us 0.2% Ni 1.8% In 38.7% Id pages
| | | | | | | | | | |
========>>>>>>>>>>>>>>>>>>>>>>% 12 forks
2 fkppw
memory totals (in kB) 2683 Interrupts 2 fksvm
real virtual free soft serial pwait
Active 995664 995664 193528 20 soft net relck
All 1338280 1338280 2241776 12 soft clock rlkok
1201 cpu0 clock noram
Namei Sys-cache Proc-cache 51 cpu0 device 104 ndcpy
Calls hits % hits % 35 cpu0 ipi 23 fltcp
560 548 98 2 0 microset ipi 360 zfod
22 shootdown ipi 54 cow
Disks: seeks xfers bytes %busy 12 cpu0 imb ipi 32 fmin
ld0 2 15K 4.0 synch fpu ipi 42 ftarg
ld1 3 46K 3.9 discard fpu ipi itarg
ld2 3 46K 4.0 1197 cpu1 clock 464 wired
ld3 41 cpu1 ipi pdfre
fd0 1 microset ipi pdscn
cd0 26 shootdown ipi
sd0 1 15K 3.6 14 cpu1 imb ipi
sd1 5 77K 3.8 synch fpu ipi
sd2 discard fpu ipi
sd3 1 15K 3.6 40 kn300 irq 12
md0 3 kn300 irq 16
md1 kn300 irq 36
ccd0 7 107K 3.8 8 kn300 irq 44
isa irq 4
Now if we got rid of that ever-empty "seeks" column then we could
probably stuff in the "filepages", "execpages", "anonpages",
"nbufpages", and "freepages" counts and percentages.....
--
Greg A. Woods
+1 416 218-0098 VE3TCP RoboHack <woods@robohack.ca>
Planix, Inc. <woods@planix.com> Secrets of the Weird <woods@weird.com>