Hi Ionel & Malte, we are creating this issue to bring to your attention the issues we are running into while implementing “xor” flow construct for pod-to-pod anti-affinity functionality.
As regards to “and” flow network construct, it seems current min-cost flow solver needs to be enhanced to generalized min-cost flow network based solver. We are not sure at this time how to go about accomplishing all this.
In the meanwhile, we are addressing pod-to-pod affinity/anti-affinity complex constraints by processing one pod at a time using multi-scheduling rounds, as suggested in Malte’s thesis.
Let us know if there is a way to address issues we are running into while implementing “xor” and “and” flow constructs for solving affinity/anti-affinity complex constraints problem. Following is the Google Doc link for all the issues we have encountered so far. Thanks.
Hi Ionel & Malte, we are creating this issue to bring to your attention the issues we are running into while implementing “xor” flow construct for pod-to-pod anti-affinity functionality.
As regards to “and” flow network construct, it seems current min-cost flow solver needs to be enhanced to generalized min-cost flow network based solver. We are not sure at this time how to go about accomplishing all this.
In the meanwhile, we are addressing pod-to-pod affinity/anti-affinity complex constraints by processing one pod at a time using multi-scheduling rounds, as suggested in Malte’s thesis.
Let us know if there is a way to address issues we are running into while implementing “xor” and “and” flow constructs for solving affinity/anti-affinity complex constraints problem. Following is the Google Doc link for all the issues we have encountered so far. Thanks.
https://docs.google.com/document/d/1kdiDXiLJ2glJ35AWRMX2H3YMOznNgkNu8-4sG4z9HRo/edit?usp=sharing
@ms705 @ICGog @shivramsrivastava