A Performance Fluctuation-Aware Stochastic Scheduling Mechanism for Workflow Applications in Cloud Environment

Fang DONG  Junzhou LUO  Bo LIU  

Publication
IEICE TRANSACTIONS on Information and Systems   Vol.E97-D   No.10   pp.2641-2651
Publication Date: 2014/10/01
Online ISSN: 1745-1361
DOI: 10.1587/transinf.2013THP0016
Type of Manuscript: Special Section PAPER (Special Section on Frontiers of Internet of Things)
Category: 
Keyword: 
cloud computing,  virtual machine,  performance fluctuations,  workflow scheduling,  

Full Text: PDF>>
Buy this Article




Summary: 
Cloud computing, a novel distributed paradigm to provide powerful computing capabilities, is usually adopted by developers and researchers to execute complicated IoT applications such as complex workflows. In this scenario, it is fundamentally important to make an effective and efficient workflow application scheduling and execution by fully utilizing the advantages of the cloud (as virtualization and elastic services). However, in the current stage, there is relatively few research for workflow scheduling in cloud environment, where they usually just bring the traditional methods directly into cloud. Without considering the features of cloud, it may raise two kinds of problems: (1) The traditional methods mainly focus on static resource provision, which will cause the waste of resources; (2) They usually ignore the performance fluctuation of virtual machines on the physical machines, therefore it will lead to the estimation error of task execution time. To address these problems, a novel mechanism which can estimate the probability distribution of subtask execution time based on background VM load series over physical machines is proposed. An elastic performance fluctuations-aware stochastic scheduling algorithm is introduced in this paper. The experiments show that our proposed algorithm can outperform the existing algorithms in several metrics and can relieve the influence of performance fluctuations brought by the dynamic nature of cloud.