Dear experts,

how can I implement a load balancer like this in current linux scheduler?

/*
Initial state: both cores are switched off
Filling cores: applies when there is a ready task T
Step1: Is there any core empty?
If so, if core A is empty then launch T to core A,
otherwise launch T to core B
If not, go to step 2.
Step2: Launch T to the core less loaded,
If both cores are equally loaded then
increase frequency and launch T to core A.
Reducing frequency: applies when a task T finishes
Step3: Are both cores equally loaded?
If so, reduce frequency

*/

Suggestions please

Why do you want to? Do your tasks require this, have you found no existing scheduler configurable to suit your needs, or is it just to learn how to do it?

What specific hardware is this targetted for? You seem to assume dual-core rather than generic multi-core and the availability of variable(stepped?) frequency.

I'm no expert on this, but just having an OS running on a device is probably going to mean there's several user space processes already running (if not kernel ones), so how do you define equally loaded? I don't believe this to be an easy field to step into and improve over what's already out there.

for academic purpose.. doing a project

In that case I suggest you consult your material as I am sure the idea of a project is for you to learn how as opposed to someone telling you.

0 members found this post helpful.