Solaris

I want to know how can in Solaris i can:
>Set Process priority to maximum
>Set Process Affinity.

Linke in windows task manager we can set process priority to following levels:
Low, Below Normal,Normal , Above Normal , High , Realtime
And we can set process affinity to run on all processors.

I am also interested to know how can i check process affinity in Solaris?
> Means on which processors , a process is running?

pbind sets "affinity" - in Solaris it is called binding
psradm -F undoes binding

renice lets you change process priority of an existing process. BE CAREFUL. Do not use real time priorities or the system can become unusable. AND. Be sure to read the "nice man pages" - nice is a command and it explains what thpriority numbers mean. They are backwards from what you think now, I'm sure.

"psradm -F" undoes the binding when a CPU is put in off-line or spare state but I wouldn't recommend doing it that way.

To bind a process:

Code:

pbind -b cpu pid

To undo this binding:

Code:

pbind -u pid

These kind of binds are shared, i.e. the affected CPUs are still usable by other processes.

You can also create sets of cpus with psrset and bind/unbind processes to theses sets. Unlike those create with pbind, these bindings are exclusive meaning only bounded processes can use the CPUs in a set.

A better question would be what problem are you trying to sovle?

Binding to a specific processor almost certainly isn't going to gain you much if any performance, and even then it's only likely to gain you performance if you're running on a NUMA machine and have taken steps to ensure the memory used by your process is local to the CPU(s) you bind to.

And the only way raising the priority of the process is going to help is if it's fighting with other processes for CPU time. If you process is already getting 100% CPU utilization or close to it (use prstat), upping the priority won't help a bit.

If many processes are fighting for CPUs, exclusive binding to a processor set might still help.
A single process can be efficiently prioritized when the fair share scheduler is used instead of the usual time sharing one. This is certainly a better approach than processor binding.

Thanks everyone
I have a new Question in same context

I can increase process priority with nice command but still it takes too much time
And CPU is still 93% idle.
CPU states: 93.6% idle, 5.9% user, 0.3% kernel, 0.3% iowait, 0.0% swap
Memory: 32.0G real, 17.6G free, 9.5G swap in use, 17.0G swap free

PID USERNAME THR PR NCE SIZE RES STATE TIME FLTS CPU COMMAND
20995 root 1 57 -19 958M 854M cpu18 10:34 0 5.91% sort

I binded it to one cpu22 but still it is taking alot of time
how can i also increse PR value?
PID USERNAME THR PR NCE SIZE RES STATE TIME FLTS CPU COMMAND
20995 root 1 59 -19 1.3G 853M cpu22 14:34 0 5.97% sort

If your process tops at 6% CPU while there is still CPU available, there is nothing you can do using CPU binding or process priorities. Your process is just not enough multi-threaded to better make use of your CPUs.