A multi-tenant hierarchical modeling for cloud computing workload. Issue 4 (1st October 2016)

• Record Type: Journal Article
• Authors: An, Chunyan; Zhou, Jiantao; Liu, Shuai; Geihs, Kurt
• Is Part Of: Intelligent automation & soft computing. Volume 22:Issue 4(2016)
• Page Start: 579
• Page End: 586
• Publication Date: 2016-10-01
• Subjects: Cloud computing -- hierarchical workload model -- workload generator; Artificial intelligence -- Periodicals; Intelligent control systems -- Periodicals; 003.5
• DOI: 10.1080/10798587.2016.1152774
• Languages: English
• ISSNs: 1079-8587

Abstract

Abstract: Diverse realistic workloads are urgently needed by SaaS providers and researchers to study the cloud environment. Workload model is an important way to specify and produce workloads. However, existing workload models are usually simplified. They cannot comprehensively describe the variations of the behaviors in user-level, application-level and service level in real environment. Moreover, existing workload models are aimed to only specific applications. They fail to support multi-users, multi-applications and multi-service-units in one workload model. To solve the problem, a hierarchical workload modeling approach is proposed in this paper. The approach constructs a cloud workload model on three layers: User, application, service-units. At each layer two key time-varying variations are captured; amount and composition. Then the workload model is the superposition of the three layers. The approach provides a unified model for different applications and various user behaviors. The experiments give the hierarchical workload models of three examples, which applications are Bag-of-Tasks, MapReduce and self-contained service-units. The experimental results show the flexibility of the novel workload modeling. Finally, architecture of a workload generator based on the model is given.