Scholarlink

The Lyapunov drift-based scheduling algorithms control the queue length to stabilize the system and maximize the throughput for multi-resource multi-queue cloud/edge systems. However, these algorithms often ignore the heterogeneity in job length, thus leading to a long delay for job processing. To this end, this paper takes the job completion time (JCT) as a penalty in the drift-plus-penalty expression to minimize the average job competition time, and proposes a Lyapunov drift-plus-time (LDPT) based scheduling algorithm for multi-resource queuing computing systems. The optimality of LDPT with respect to average JCT and queue length is analyzed. By using the Amazon EC2 to simulate, the results show that LDPT can improve average JCT and average queuing time fairness.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Minimizing Average Job Completion Time for Multi-resource Queuing Computing Systems Using Drift-Plus-Penalty

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Chuanxu Chen,
  • Ruiting Li,
  • Quansheng Guan,
  • Zhan Shi,
  • Ying Zeng,
  • Jianping Zheng

摘要

The Lyapunov drift-based scheduling algorithms control the queue length to stabilize the system and maximize the throughput for multi-resource multi-queue cloud/edge systems. However, these algorithms often ignore the heterogeneity in job length, thus leading to a long delay for job processing. To this end, this paper takes the job completion time (JCT) as a penalty in the drift-plus-penalty expression to minimize the average job competition time, and proposes a Lyapunov drift-plus-time (LDPT) based scheduling algorithm for multi-resource queuing computing systems. The optimality of LDPT with respect to average JCT and queue length is analyzed. By using the Amazon EC2 to simulate, the results show that LDPT can improve average JCT and average queuing time fairness.