Heuristic Load Scheduling Algorithm for Stateful Cloud BPM Engine

Abstract

With the increasing popularity of cloud computing technology, the traditional Business Process Management System (BPMS) begins to transform to the architecture that deployed in the cloud. Since the traditional BPMS is often implemented as a stateful single-instance architecture, the cloud BPMS providers will encounter the stateful service load scheduling problem when they refactor the traditional BPMS to the microservice architecture that deployed on the cloud server. In order to help realize the transformation of traditional BPMS and improve the load capacity of cloud BPMS with limited computing resources, we propose a heuristic load scheduling algorithm for stateful service scheduling. The algorithm makes use of the busyness metrics of the single BPMS engine instance microservice in the cloud BPMS architecture. Because the resource scheduling problem is always defined as online bin packing problem, we improve the Best-Fit algorithm to solve this kind of problem. We come up with the Best-Fit Decreasing algorithm based on cloud BPMS engine busyness measuring and load prediction to schedule the computing resources to business process instances. Compared to some common load scheduling algorithms, our algorithm help cloud BPMS increase load level by at least 15%.