Elastic and flexible multi-stage task scheduling with deadline-constraint in clouds
Jie Zhu, Xiaoping Li · 2016
Cloud has become an attractive computing platform which offers seemly unlimited and computing resources to public. From perspective of data centers which offer cloud services, however, computing resources are limited and operating cost restricts cloud service quality. In order to balance between cost and service quality, the scheduling module, as the core component of the management system of data centers, should be able to sophisticatedly schedule computing jobs with high utilization of computing resources. In the paper, we present efficient heuristics for the scheduling module to yield elastic and flexible schedule plans for desired service quality with less cost, which can automatically scale up/down computing instances in response to workload over time. Computing jobs are specified as flowshop type jobs, which are multi-stage tasks with linear processing routes. Jobs are assigned hard deadlines according to service quality. The goal is to ensure all jobs are finished within their deadlines with the minimum number of elastic computing instances. Experimental results show that the proposed heuristics can effectively improve utilization of computing resources and guarantee cloud service quality.