Scheduling Parallel Soft Real-Time VM in Dynamic Workloads. Issue 2 (2nd April 2016)

• Record Type: Journal Article
• Authors: Ding, Xiaobo; Ma, Zhong; Dai, Xinfa
• Is Part Of: Intelligent automation & soft computing. Volume 22:Issue 2(2016)
• Page Start: 281
• Page End: 287
• Publication Date: 2016-04-02
• Subjects: Concurrent program -- soft real-time -- VM -- synchronism scheduling; Artificial intelligence -- Periodicals; Intelligent control systems -- Periodicals; 003.5
• DOI: 10.1080/10798587.2015.1095478
• Languages: English
• ISSNs: 1079-8587

Abstract

Abstract: Virtual machine (VM) is widely used in many fields now. The CPUs of VM, different from those of physical machines, cannot be ensured to be on line at the same time, and the number of VMs will affect the performance of the parallel VM. So concurrent programs working on the VM is not efficient, and the soft real-time concurrent applications would be invalid especially in the worst case. Based on Xen Credit scheduler, this paper analyzes the factors affecting the performance of Parallel Soft Real-Time task VM (PSRTVM) and raises a Parallel Soft Real-Time Scheduling Algorithm (PSRTSA), which could ensure the performance of PSRTVM in dynamic workloads. In PSRTSA, the synchronized scheduling is proposed to meet the parallel workloads, and the opportunity of synchronized scheduling is set to meet the delay of the worst case. The percentage pre-allocation of CPU time is used to ensure the resource of PSRTVM in dynamic workloads of system. The evaluation shows the improvement of the PSRTSA compared to the default Credit scheduler and RT-Credit scheduler based on scheduling synchronization.