UNIX for Advanced & Expert Users

I have a requirement where i need to restrict users into reusing the first 3 exec machine they are using so as to reduce license usage of our tools.

So, lets say that at any instant an user is using 3 machine - mac1,mac2,mac3. Now if he is going to run a new job, he is expected to re-use one of the three machine already in use.

Up to this point , I hope my requirement is clear. Please prompt me if you have any doubt

Now, I aim to design an algorithm which decides which is the best machine to use. I need following input to design the algo:-

1. Among mac1,2,3 , what parameters can be used to decide what machine is best to exec a job.

There are few other requirements as well, but i would like to try and fail before posting them here. The one posted above is the biggest bottleneck i have

---------- Post updated at 04:42 PM ---------- Previous update was at 04:41 PM ----------

I went through lsload command, but frankly I need some sort of guidelines to decide what load factors/user logins/memory/etc signals health of a exec machine.

Assume you know the load profile your jobs generate e.g., low cpu, lots of I/O, some network traffic.

In this case I/O would be the determining factor. When you have no clue what some job will consume you try to balance cpu usage. Sometimes load balancing larger jobs means waiting for resources. If you know that job is a hog ahead of time.

The reason load balancing network site requests is usually attainable without too many issues -- they mostly have similar profiles, some db access, a small amount of I/O.

Since you are asking an extremely general question, you got an extremely general answer. What you are doing is playing "OS" in a sense - you are doing what the OS does - try to schedule jobs. In such a way that "everybody is happy". There are round-robin schedulers, fair share schedulers - as examples. google for each of those to see what is entailed.-----