Hello
My application consists of several hundreds of threads.
I want to create a loadcontrol mechanism at application level so that
when the cpuload of a group of threads crosses a certain
(configurable) threshold, the threads will voluntary yield until the
load condition is over. For this I need to have a mechanism int the
kernel that calculates the consolidated load of each of the
thread-groups.
Is such a mechanism in place in the kernel?
I could for example define a cgroup per group I want to define and
assign the threads to the correct group. I have no clue whether
internally in the kernel the total CPUload consumed per cgroup is
kept. If that would be the case, it would be ideal.
I do not want to make the sum of all the thread's load (for example
based on the info in the proc filesystem) because this is generating
too much overhead.
Any hint/help/info is much appreciated.
Best regards,
Ronny