Model-Based Comparison of Cloud-Edge Computing Resource Allocation Policies

Abstract

Abstract The rapid and widespread adoption of internet of things-related services advances the development of the cloud-edge framework, including multiple cloud datacenters (CDCs) and edge micro-datacenters (EDCs). This paper aims to apply analytical modeling techniques to assess the effectiveness of cloud-edge computing resource allocation policies from the perspective of improving the performance of cloud-edge service. We focus on two types of physical device (PD)-allocation policies that define how to select a PD from a CDC/EDC for service provision. The first is randomly selecting a PD, denoted as RandAvail. The other is denoted as SEQ, in which an available idle PD is selected to serve client requests only after the waiting queues of all busy PDs are full. We first present the models in the case of an On–Off request arrival process and verify the approximate accuracy of the proposed models through simulations. Then, we apply analytical models for comparing RandAvail and SEQ policies, in terms of request rejection probability and mean response time, under various system parameter settings.