Joint Jobs Scheduling and Lightpath Provisioning in Fog Computing Micro Datacenter Networks

Abstract

With the growth in data volume and latencysensitive applications, fog computing micro datacenters (FC-mDC) have been proposed as a promising solution. The FC-mDC locally processes raw data requested by the computation task. A group of related parallel computation tasks is called a job, and the required raw data for a job request are mostly stored in multiple FC-mDCs. When establishing multiple job requests in the FC-mDC network, effectively managing job scheduling and data transmission is an open challenge. In this paper, a joint optimization algorithm of multiple jobs scheduling and lightpath provisioning (MJSLP) is investigated for minimizing average completion time in FC-mDC elastic optical networks. The MJSLP algorithm includes two key submethods that have a progressive relationship: first, for the minimization of a single job completion time (i.e., the slowest task completion time in a job), a mathematical model and a heuristic algorithm are developed. Then, on the basis of single job completion time, an integer linear programming formulation and a heuristic algorithm of routing, modulation level, and frequency slots are proposed. Simulation results show that the MJSLPalgorithmoutperforms the scheduling-only and routing-only algorithms in terms of average completion time and frequency slots.