Tek-Tips is the largest IT community on the Internet today!
Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!
-
Congratulations Mike Lewis on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!
Steps to Check for High CPU usage...
- Thread starter AIXdream
- Start date
As TSch said, I like topas so much and i use it always to monitor my system
But there are alot of CPU monitoring commands that you can use as well
Like
sar, vmstat, ps aux
Check this link for more information on aix performance tools:
Regards,
Khalid
But there are alot of CPU monitoring commands that you can use as well
Like
sar, vmstat, ps aux
Check this link for more information on aix performance tools:
Regards,
Khalid
hi ,
This is not a straight forward task , but as a guidance what you can do is check the following :-
1. You need to check this over a period of time , and then determine if slow performance is at particular times of the day , or just particular days etcc... also it would be a good to obtain some information which you can use as bencmark test , ideally when the system may be for certain times of the day , running without any problem you can check this
1. run vmstat say every minute for 24 hours preferabbly longer
look at the cs parameter i.e. context switching , a high value here indicates that a thread is taken off a CPU and other one running on that CPU on a regular occurance this could e a thread has taken its full time allocated to run its task and then another one has started , or a thread is gone to sleep waiting for a resource or a lock ) value should be low
the other column r is the runnabe threads , the rule of thumb here is the r value divided by number of CPUS if greater than one , there could be a bottle neck
also us+sy ( add together) is 100% there may be CPU bottleneck
check these values as a starting point
This is not a straight forward task , but as a guidance what you can do is check the following :-
1. You need to check this over a period of time , and then determine if slow performance is at particular times of the day , or just particular days etcc... also it would be a good to obtain some information which you can use as bencmark test , ideally when the system may be for certain times of the day , running without any problem you can check this
1. run vmstat say every minute for 24 hours preferabbly longer
look at the cs parameter i.e. context switching , a high value here indicates that a thread is taken off a CPU and other one running on that CPU on a regular occurance this could e a thread has taken its full time allocated to run its task and then another one has started , or a thread is gone to sleep waiting for a resource or a lock ) value should be low
the other column r is the runnabe threads , the rule of thumb here is the r value divided by number of CPUS if greater than one , there could be a bottle neck
also us+sy ( add together) is 100% there may be CPU bottleneck
check these values as a starting point
AIXdream.
What tools is telling you "High cpu usage on server AIX 5.2". Is this real time output or a report from previous day?
If this is real time then I would use tprof since its does a trace of the system for high cpu users
syntax? first cd to an empty dir and then run
#tprof -x sleep 30
then tprof will name a fie.
For cpu-bound systems;there are limited resolutions:
1) a process is a run-away and can be killed then the application owner needs to have the code examinied for a defect
2) an application user kicked off a job that should be scheduled for a time with fewer users on the system
3) There server does not have enough processors for the workload that you are giving it so add more processors OR
remove some workload.
4) Besides tuning applications so that they are more efficient there is not system tuning for a system that has either too many processes to service OR one or two cpu-hogs
Hope that this helps
What tools is telling you "High cpu usage on server AIX 5.2". Is this real time output or a report from previous day?
If this is real time then I would use tprof since its does a trace of the system for high cpu users
syntax? first cd to an empty dir and then run
#tprof -x sleep 30
then tprof will name a fie.
For cpu-bound systems;there are limited resolutions:
1) a process is a run-away and can be killed then the application owner needs to have the code examinied for a defect
2) an application user kicked off a job that should be scheduled for a time with fewer users on the system
3) There server does not have enough processors for the workload that you are giving it so add more processors OR
remove some workload.
4) Besides tuning applications so that they are more efficient there is not system tuning for a system that has either too many processes to service OR one or two cpu-hogs
Hope that this helps
- Status