Pod allocation algorithms to handle time-varying numbers of applications occurrence on the Access-Metro edge computing

Abstract

AMec (Access-Metro edge computing) is a technology that can reduce the number of edge servers by utilizing in-network excess computing resources such as unallocated cores of control CPU of network devices as a node of the AMec cluster. The important issue is service interruption in that computing resources become unavailable due to executing their main job (e.g. device control and management jobs). In this paper, to avoid interruptions, two types of pod allocation algorithms (both static and dynamic thresholds) are evaluated. The dynamic threshold method is effective in keeping a high allocation success rate even for time-varying numbers of application occurrences.