Workflow-based resource allocation to optimize overall performance of composite services

  • Authors:

  • BangYu Wu;Chi-Hung Chi;Zhe Chen;Ming Gu;JiaGuang Sun

  • Affiliations:

  • School of Software, Tsinghua University, Beijing, 100084, China and Key Laboratory for Information System Security, Ministry of Education of China, Beijing, 100084, China;School of Software, Tsinghua University, Beijing, 100084, China and Key Laboratory for Information System Security, Ministry of Education of China, Beijing, 100084, China;School of Software, Tsinghua University, Beijing, 100084, China and Key Laboratory for Information System Security, Ministry of Education of China, Beijing, 100084, China;School of Software, Tsinghua University, Beijing, 100084, China and Key Laboratory for Information System Security, Ministry of Education of China, Beijing, 100084, China;School of Software, Tsinghua University, Beijing, 100084, China and Key Laboratory for Information System Security, Ministry of Education of China, Beijing, 100084, China

  • Venue:
  • Future Generation Computer Systems
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

In software service provision, the overall performance of a composite service is often the ultimate focus of concern rather than those of its individual components. This opens new opportunities for resource allocation because with its service workflow definition, more accurate prediction of its individual components' dynamic workload is possible, thus resulting in better utilization of resources. In this paper, we propose to improve resource allocation through tracing and prediction of workload dynamics of component services as requests traverse and pipeline through the workflow. Factors affecting service workload such as service time, transition probability, replication overhead for additional service etc. as well as the uncertainty in request arrival time are all taken into consideration in our model. The goal is to maximize the number of requests completed under the constraints of limited available resources. Experimental study on TPC-W and synthetic workflow shows that our dynamic workflow-based resource allocation scheme is much more efficient in enhancing the overall performance of composite services than current resource allocation schemes do.