server performance..

Hi,

would like to know the server status from the following 'top' out put.  Because, the application is giving a bad performance. would like to know whether the load is  within the acceptable limit.

Murali...

System: shpu28                                        Tue Feb  3 10:03:31 2004
Load averages: 0.67, 0.64, 0.62
211 processes: 148 sleeping, 62 running, 1 stopped
Cpu states:
CPU   LOAD   USER   NICE    SYS   IDLE  BLOCK  SWAIT   INTR   SSYS
 0    0.88  50.2%   0.0%   7.6%  42.2%   0.0%   0.0%   0.0%   0.0%
 1    0.44  54.8%   0.0%  16.7%  28.5%   0.0%   0.0%   0.0%   0.0%
 2    0.63  41.2%   0.0%  21.3%  37.5%   0.0%   0.0%   0.0%   0.0%
 3    0.73  43.2%   0.0%  41.2%  15.5%   0.0%   0.0%   0.0%   0.0%
---   ----  -----  -----  -----  -----  -----  -----  -----  -----
avg   0.67  47.4%   0.0%  21.7%  30.9%   0.0%   0.0%   0.0%   0.0%

Memory: 1287356K (761728K) real, 1770636K (1039216K) virtual, 2826912K free  Pag
e# 1/24

CPU TTY  PID USERNAME PRI NI   SIZE    RES STATE    TIME %WCPU  %CPU COMMAND
 1   ? 26343 root     241 20  7924K  1652K run   1289:32 98.69 98.52 ndmcmgr
 0   ? 21057 root     241 20  4844K   324K run   1288:58 98.53 98.36 cdstatm
 1   ?  6695 uora7ap  152 20   855M 17488K run      1:40 23.76 23.72 oraclejpfl1
 2   ?  2785 root     -16 20 42056K 10616K run    107:33  5.82  5.81 midaemon
 3   ?  2425 root     152 20   416K   416K run    241:12  3.08  3.07 lltd
 2   ? 10545 uora7ap  152 20   843M  4176K run      0:02  2.16  2.15 oraclejpfl1
 0   ?    39 root     152 20  5120K  5120K run     47:09  1.64  1.63 vxfsd
 1   ?  6799 uora7ap  152 20   847M  9164K run      4:36  0.73  0.73 ora_lms0_j

The loads on the cpus are all very reasonable. Most likely you are disk bound.

Hi,

    I thought load should always be less than 0.3 for the better performance. And one more thing is, CPU utilization is 99% for the processes ndmcmgr & cdstatm. wudn't this impact the response of the other processes...

Murali

I don't know where you got that .3 but it's way too low. You have double digit idle times on every cpu. You do have two processes that spend 98% of their time in the cpu. The kernel knows this and has given them a nasty priority. They only run when nothing else wants to. Since you have double digit idle times on each of your four cpu's, yes they run a lot.

HP has a product called glance or maybe it's glance-plus now. It looks like you can download a trial version here. Try it. It will tell where your bottlenecks are.

Hi,

Thanks for the info.. was quite informative...

Murali