Non-Critical:Quantifying the effect of change in "autoup" value.
From: sreenath sarikonda (sreeredhat_at_yahoo.com)
Date: 10/28/04
- Previous message: John skreen: "Creating additional swap & separating /tmp"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Thu, 28 Oct 2004 13:54:47 -0700 (PDT) To: sunmanagers@sunmanagers.org
Hello Gurus,
We host a financial application(something like
portfolio management software or investment accounting
system). We install one instance of the application
per client with a specific user id.So whenever the
clients are using this system the engine(user) uses
atleast 40% of cpu and when it quartely reporting this
process may run for 5 hrs.When two clients use it at
same time one may use up 50% of cpu and the other may
use up 40% of cpu. And if we see the vmstat output the
run lenght queue hits very high(20)on a 4 cpu system
example. This happens even one client(user) uses
90% of cpu. Edited output for readability.There is lot
of free memory in system.
r b w in sy cs us sy id
2 0 0 753 740 423 15 3 82
36 0 0 4705 77836 81814 93 7 0
34 1 0 3483 47341 47931 97 3 0
34 0 0 2271 15493 15913 93 7 0
34 0 0 3418 17419 17964 97 3 0
I was reading the SPEC2000 benchmarks for SunFire
systems and they optimized setting the
autoup=900
tune_t_fsflushr=1
Do I get any benefit from changing this value interms
of performance specific to our application.We have
systems ranging from 4Gb ot 16Gb memory.
I have a flexibility to change these values on a
development system. But there wont be any feed back on
performance increase.
What should I be looking for to find any performance
increase.
What kind of system variables I should be changing
specific to this application.
Also which model of sun is best for this kind of
application. As of now its running on a v440.
Also could someone comment on these statistics. These
statistics represent samples(in one month) of
iostat,of a disk on which application is installed,
taken only when system is atleast
30% busy in cpu.This is a cpu intensive appllication
and there is no disk i/o. However there is high
asvc_t. Could this be attributed to that the i/o
request was not processed by cpu as
system is busy. Please give me additional comments
which will give greater insight into how to look
iostat output edited.Also disk under consideration is
mirrored with SDS.
kr/s kw/s wait actv wsvc_t asvc_t %w %b
4.6 12.3 0.0 0.1 7.3 60.9 1 1
0.0 16.0 0.1 0.0 26.9 6.8 5 1
4.6 12.3 0.0 0.1 7.3 60.9 1 1
0.0 11.0 0.4 0.1 128.0 21.4 21 3
0.0 8.0 0.0 0.1 0.0 42.6 0 5
4.6 12.3 0.0 0.1 7.3 60.9 1 1
0.0 0.0 0.1 0.0 0.0 0.0 7 0
0.0 8.0 0.2 0.0 239.6 7.3 24 1
4.6 12.3 0.0 0.1 7.3 60.8 1 1
0.0 47.9 0.0 0.0 0.0 5.7 0 2
0.0 48.1 0.4 0.0 106.9 9.1 22 3
0.0 16.0 0.0 0.0 0.0 6.3 0 1
4.6 12.3 0.0 0.1 7.3 60.8 1 1
0.0 8.0 0.0 0.0 0.6 0.6 0 1
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
- Previous message: John skreen: "Creating additional swap & separating /tmp"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
|
