DISTRIBUTED NETWORK RESOURCE ALLOCATION WITH INTEGER CONSTRAINTS

This paper considers the resource allocation problem defined over a hybrid data center and edge server network, where the allocations are subject to integer constraints, taking the granularity of network resources and service requests into account. We develop two efficient heuristic algorithms to solve this nonconvex program, both based the alternating direction method of multipliers (ADMM) and a distributed integral projection scheme. The first algorithm ignores the integer constraints and solves the relaxed convex program, while the second algorithm includes the integer constraints in the optimization process. We develop a distributed integral projection scheme, which approximately projects the resulting resource allocation strategies onto the feasible set. Numerical experiments validate the effectiveness of the proposed algorithms.