I would like to study the interaction between two applications on a CPU core, one is persistent application (NVM resident) and another is regular (DRAM resident). For this I want to only schedule these 2 applications on a core and nothing else. I am looking at Linux scheduler to accomplish this. Can someone please help me with a direction to achieve this? Can I achieve this using sched or do I need to modify the scheduler code of kernel so that scheduler does not schedule applications to the core of my interest.
2 Answers
You can use the isolcpus command-line parameter of the kernel:
This option can be used to specify one or more CPUs to isolate from the general SMP balancing and scheduling algorithms. You can move a process onto or off an "isolated" CPU via the CPU affinity syscalls or cpuset.
isolcpus is a solution but the linux kernel documentation declares it as deprecated:
isolcpus= [KNL,SMP,ISOL] Isolate a given set of CPUs from disturbance.
[Deprecated - use cpusets instead]
The cpuset sub-system of the cgroups is preferable. A tool like partrt is based on this principle to divide the CPU cores in two subsets: nrt where all the processes run and rt into which you can move your applications. Hence, the applications are isolated on some CPU cores.
Related