Worst-Case Latency Performance Of Load Balancing Through Distributed Waterfilling Algorithm

Abstract

Intelligent-meshed mobile edge computing (IMMEC) network puts great emphasis on providing low latency services. This naturally requires using fast-timescale load balancing to avoid excessive delay resulted from overloading any particular network node. One natural candidate solution for such load balancing is distributed waterfilling algorithm. In this paper, we analyze its performance by comparing it against an ideal centralized version. It is shown that the worst-case average total latency difference between the two algorithms grows linearly with the size of the network and the propagation delay among different nodes.