Research on Elastic Job Scheduling Model and Algorithm of Cloud Computing based on Hadoop

Shi Hengliang Bai Guangyi 1,3 Tang Zhenmin 1. Computer School of Nanjing University of Science & Technology, [email protected]; 2. Electronics and Information School of Henan University of Science & Technology, 3. Noah IT Solution (Suzhou)Co., LTD. Abstract A poor job scheduling strategy in cloud computing can result in lower scheduling efficiency, unbalance workload, frequent scheduling, even as thrashing machines, which cripple the cluster’s computational power. After researching many “Time for quality” job scheduling models and algorithms, the authors propose a novel elastic, positive, feedback job scheduling model ------EPSM(Elastic Positive Scheduling Model) which implements “Space for quality” strategy, making a milestone among job scheduling fields. Simultaneously, this model breaks down Mapreduce’s original architecture drawback that slave node should be assigned job passively by master node. By comparing experimental tests, this model is proved to own many advantages such as flexibility, simplicity, high throughput and short response time. The design idea is valuable to other researchers.