基于集聚系数的工作流切片与多云优化调度

Clustering Coefficient-Based Workflow Slicing and Multi-Cloud Scheduling

摘要:

已有的工作流云调度研究，通常将任务和云资源一一对应，难以解决由于频繁的数据通信而带来的完工时间上升、成本增加以及可能的故障风险等问题。因此，为减轻任务间数据通信对完工时间和成本的影响，提出了一种基于集聚系数的工作流切片与多云优化调度解决方案。通过聚类算法对工作流进行初步切片，引入集聚系数来判断和优化切片效果，并在寻找调度方案的过程中根据云实例的实际情况动态地调整切片结果。实验结果表明，所提方案能够有效地减少工作流中因大量数据通信而带来的高昂成本和完工时间。

Abstract:

Workflow scheduling in multi-cloud environment is a research hotspot and challenge in recent years. The dependencies in workflow are usually represented by the transmission of data， which also determines the execution order of tasks. Existing studies for workflow scheduling usually map each task to a different cloud resource， which is difficult to solve the problems of increasing make-span and cost， and the possible failure risk caused by frequent data communication. In order to reduce the impact of data communication between tasks， this paper proposes a workflow slicing and multi-cloud scheduling solution based on clustering coefficient. Preliminary slicing of workflow is conducted by using a clustering algorithm， and the clustering coefficient is introduced to evaluate and optimize the slicing effect. In the process of finding the optimal scheduling solution， the slicing result is adjusted dynamically according to the actual situation of cloud instances. Experimental results show that the proposed method can effectively reduce the high cost and make-span caused by large amount of data communications in workflow.