- Pod Affinity/Anti-Affinity optimization.
To enable the new CPU-MEM cost model --flow_scheduling_cost_model
flag should be set to '10'
like shown below
--flow_scheduling_cost_model=10
To enable the Pod Affinity/Anti-Affinity symmetry --pod_affinity_antiaffinity_symmetry
flag should be set to 'true' in the firmament config file
like show below
firmament_scheduler_cpu_mem.cfg=true
This fix is based above the CPU-MEM cost model. This feature currently are not supported in other cost model.