A service-oriented priority-based resource scheduling scheme for virtualized utility computing

  • Authors:

  • Ying Song;Yaqiong Li;Hui Wang;Yufang Zhang;Binquan Feng;Hongyong Zang;Yuzhong Sun

  • Affiliations:

  • Key Lab. of Computer System and Architecture, Inst. of Computing Techn., Chinese Academy of Sciences, China and National Research Center for Intelligent Computing Systems, ICT, China and Graduate ...;Key Lab. of Computer System and Architecture, Inst. of Computing Techn., Chinese Academy of Sciences, China and National Research Center for Intelligent Computing Systems, ICT, China and Graduate ...;Key Laboratory of Computer System and Architecture, Institute of Computing Technology, Chinese Academy of Sciences, China and National Research Center for Intelligent Computing Systems, ICT, China;Key Lab. of Computer System and Architecture, Inst. of Computing Techn., Chinese Academy of Sciences, China and National Research Center for Intelligent Computing Systems, ICT, China and Graduate ...;Key Lab. of Computer System and Architecture, Inst. of Computing Techn., Chinese Academy of Sciences, China and National Research Center for Intelligent Computing Systems, ICT, China and Graduate ...;Key Lab. of Computer System and Architecture, Inst. of Computing Techn., Chinese Academy of Sciences, China and National Research Center for Intelligent Computing Systems, ICT, China and Graduate ...;Key Laboratory of Computer System and Architecture, Institute of Computing Technology, Chinese Academy of Sciences, China and National Research Center for Intelligent Computing Systems, ICT, China

  • Venue:
  • HiPC'08 Proceedings of the 15th international conference on High performance computing
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

In order to provide high resource utilization and QoS assurance inutility computing hosting concurrently various services, this paper proposes aservice computing framework-RAINBOW for VM(Virtual Machine)-basedutility computing. In RAINBOW, we present a priority-based resourcescheduling scheme including resource flowing algorithms (RFaVM) to optimizeresource allocations amongst services. The principle of RFaVM is preferentiallyensuring performance of some critical services by degrading of others to someextent when resource competition arises. Based on our prototype, we evaluateRAINBOW and RFaVM. The experimental results show that RAINBOWwithout RFaVM provides 28%-324% improvements in service performance,and 26% higher the average CPU utilization than traditional service computingframework (TSF) in typical enterprise environment. RAINBOW with RFaVMfurther improves performance by 25%-42% for those critical services whileonly introducing up to 7% performance degradation to others, with 2%-8%more improvements in resource utilization than RAINBOW without RFaVM.