is the cpu under load ?

[Maaz]

I have a Linux Server, hosting/running SUN Microsystem SUNRAY(SUNRAY is a Terminal Server or LTSP like software), on this server 35 users login via dumb terminals.
these users just run OpenOffice, firefox and Evolution on this Linux Server(ldap+sunray)
Performance is simply/at least not bad(from the end-user point-of-view), but values are so high(vmstat output)
One more thing this server is also an nfs client i.e /home is not local its a nfs mount point.

Its a Quad Core Dual Processor(Xeon) 2.5 Ghz machine

# uname -a
Linux FATS008 2.6.5-7.244-smp #1 SMP Mon Dec 12 18:32:25 UTC 2005 x86_64 x86_64 x86_64 GNU/Linux

# cat /etc/SuSE-release
SUSE LINUX Enterprise Server 9 (x86_64)
VERSION = 9
PATCHLEVEL = 3

following are the details
# rpm -qa | grep java
java2-jre-1.4.2-129.14
java2-1.4.2-129.14
gcc-java-3.3.3-43.41

# rpm -q OpenOffice_org
OpenOffice_org-1.1.5-0.8

top
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
4919 khurram. 25 0 163m 95m 59m R 90.6 0.6 25830:02 soffice.bin
17906 shakeel. 25 0 128m 73m 52m R 90.6 0.5 9623:08 soffice.bin
25103 yousuf.a 25 0 148m 87m 54m R 89.2 0.5 5787:13 soffice.bin
27082 shakeel. 25 0 124m 68m 49m R 87.7 0.4 12362:19 soffice.bin
24350 ahmed.ab 25 0 124m 68m 49m R 80.2 0.4 66:42.84 soffice.bin
26536 sabeen.r 25 0 151m 90m 55m R 75.8 0.6 1044:40 soffice.bin
24333 ahmed.ab 25 0 124m 68m 49m R 63.9 0.4 66:47.26 soffice.bin
1029 shuja.si 16 0 89860 19m 13m R 35.7 0.1 0:02.07 krdc
11444 root 16 0 66196 60m 2608 R 16.3 0.4 1:02.77 Xnewt
9893 root 15 0 14704 9m 1968 R 8.9 0.1 0:01.11 Xnewt
1299 root 15 0 22892 3916 1956 R 8.9 0.0 0:00.13 top
5306 root 15 0 100m 5284 3732 S 4.5 0.0 0:09.97 owcimomd
8663 amjad.fa 15 0 106m 62m 21m S 3.0 0.4 8:18.70 firefox-bin
12742 shuja.si 15 0 101m 22m 16m S 3.0 0.1 0:01.86 kdeinit
13314 shuja.si 16 0 144m 90m 22m R 3.0 0.6 8:51.92 firefox-bin
5142 root 16 0 107m 99m 2476 R 1.5 0.6 2:57.14 Xnewt
5486 tanweer. 15 0 148m 30m 16m S 1.5 0.2 0:01.02 nautilus
6058 tanweer. 16 0 192m 144m 23m S 1.5 0.9 3:42.37 firefox-bin
7426 root 16 0 28116 22m 2488 R 1.5 0.1 3:18.40 Xnewt
9986 amjad.fa 15 0 106m 62m 21m S 1.5 0.4 1:44.82 firefox-bin
9987 amjad.fa 15 0 106m 62m 21m R 1.5 0.4 3:24.29 firefox-bin
9988 amjad.fa 15 0 106m 62m 21m R 1.5 0.4 3:33.69 firefox-bin
12650 shuja.si 15 0 95040 19m 14m S 1.5 0.1 0:01.81 kdeinit
12672 shuja.si 15 0 96652 22m 16m S 1.5 0.1 0:01.09 kdeinit
12963 shuja.si 15 0 92280 19m 13m S 1.5 0.1 0:00.54 kdeinit
14727 shuja.si 15 0 144m 90m 22m S 1.5 0.6 2:35.62 firefox-bin
15038 tabinda. 15 0 521m 457m 54m S 1.5 2.9 5:10.94 soffice.bin
17828 saima.is 15 0 173m 72m 29m S 1.5 0.5 0:11.51 kmail
28255 shuja.si 15 0 141m 73m 50m S 1.5 0.5 0:03.85 soffice.bin
30918 amjad.fa 15 0 229m 32m 13m S 1.5 0.2 0:43.60 java_vm
32011 syed.moh 15 0 142m 79m 51m S 1.5 0.5 0:07.21 soffice.bin

# vmstat
procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
r b swpd free buff cache si so bi bo in cs us sy id wa
12 0 129960 1376284 274916 8107672 0 0 1 6 1 2 36 44 19 0
9 1 129960 1401588 274916 8107672 0 0 0 338 2191 25944 46 49 5 0
10 0 129960 1400936 274916 8107672 0 0 0 2 1813 18697 46 48 6 0
13 0 129960 1380636 274920 8107668 0 0 0 218 2258 14180 56 44 0 0
8 0 129960 1383668 274920 8107668 0 0 0 0 2228 13322 53 46 1 0
8 0 129960 1365788 274920 8107668 0 0 0 710 2291 15146 54 45 1 0
13 0 129960 1364128 274920 8107668 0 0 0 0 2135 11203 56 44 0 0
8 0 129960 1350488 274920 8107668 0 0 0 456 1922 12112 57 43 0 0
17 0 129960 1350380 274920 8107668 0 0 0 0 1568 13266 47 46 6 0
12 0 129960 1350072 274920 8107668 0 0 0 40 1866 10472 45 47 7 0
11 0 129960 1349816 274920 8107668 0 0 0 40 2362 17051 54 45 0 0
7 0 129960 1376044 274920 8107668 0 0 0 308 2001 13409 56 43 1 0
12 0 129960 1367580 274920 8107668 0 0 0 0 1999 18149 51 47 2 0
11 0 129960 1357388 274920 8107668 0 0 0 410 1773 11169 51 48 1 0
16 0 129960 1349568 274920 8107668 0 0 0 114 2060 12658 54 46 0 0
8 0 129960 1349576 274920 8107668 0 0 0 0 1711 13843 50 46 3 0

I wana know are the above values ok ? or does the cpu is under high load ?

[Ivan Ferreira · Accepted solution]

Considering the output of vmstat (si, so, wa), your system is not having a bottleneck on memory resources or device I/O.

Your system may be CPU bound, you have a high sys time, maybe normal for a terminal server. I would suggest more CPU for your server, and if you can, use NFSv4, that is able to handle more request in a single operation to try to reduce the sys time.

Maybe, you should also tune the I/O elevator scheduler.

¿Are those top OpenOffice processes doing the right thing?

[Maaz]

Hi Ivan Ferreira, thanks for help.

> if you can, use NFSv4, that is able to handle more request in a single
> operation to try to reduce the sys time.
this machine is not a NFS server(it nfs client)... do you still suggest me to use NFSv4 ?

> Maybe, you should also tune the I/O elevator scheduler.
hmmm.... please explain .. didn't get you Sir ?

> Are those top OpenOffice processes doing the right thing?
these users have usually have 10-20 mb files(.doc, and .xls) I mean each files has the size of 10 or 20 mb, and usually these user has in b/w 4 to 7 files open at any time, but if you are asking something else, then please explain ?

[Heironimus]

You would need to run NFSv4 on both ends, so it doesn't matter if this system is the client or the server.

"I/O elevators" are the Linux name for I/O scheduling algorithms. I believe these are only present in 2.6 kernels. However, other tunables related to I/O are present both older and newer kernels.

[Steven E. Protter]

Shalom,

This system is running an enormous amount of processes. Its like dozens of PC's running on one system.

You ask the question are the values okay?

Premise is wrong. The right question is:

1) Is there a response problem? Are you collecting data bsaed on user complaints? If the answer is yes, and that is why you are collecting data then, yes you have a problem. If you have no user complaints these numbers are not out of line.

2) Don't let the horse drag the cart. Numbers are meaningless. User perception is important.

SEP

[Maaz]

Thanks Ivan and Heironimus for help


Anticipatory scheduler ("as scheduler")

The anticipatory scheduler implements request merging, the elevator and in addition optimizes for physical disks by avoiding too many head movements. It tries to solve situations where many write requests are interrupted by a few read requests. After a read operation it waits for a certain time frame for another read and doesn't switch back immediately to write requests. This scheduler is not intended to use for storage servers!

source: http://www.ibm.com/developerworks/linux/linux390/perf/tuning_rec_dasd_ioScheduler.html

I just didnt get the last line
"This scheduler is not intended to use for storage servers!"
my question is whats the Storage Server
I have a HP SAN, i have created a vdisk and present that disk to Linux Server, now which one is Storage Server
(a), HP SAN device
or
(b) Linux Server
and how can I change the I/O scheduler ?

and thanks SEP for help and valuable recommendations.
yes we have some complaints from user, but only when they have open very large file e.g 40 MB file(.xls) open, otherwise performance is at least not bad (if not super fast or good)

[Ivan Ferreira]

Please see this link:

http://www.redhat.com/magazine/008jun05/features/schedulers/