Pota: Maximizing Profit for Task-Level Scheduling for Data Center Networks

Scheduling is one key issue in Data Center Networks (DCN). Earlier research work usually focuses on flow-level scheduling, while more and more people are aware of the benefits of task-level scheduling in recent years. Most existing task-level scheduling methods schedule flows of one task together in order to reduce average completion time. However, few works discuss the efficient task-level scheduling in the view of the profit of tasks. To address this problem, we propose a novel task-level scheduling method named Pota, whose target is to maximize the profit of completing tasks within their deadline. To achieve this goal, we propose a maximizing profit optimization model, and then present an efficient scheduling 1/2-approximate algorithm. Based on the proposed algorithm, we design and implement Pota. We also conduct comprehensive experiments to evaluate the performance of Pota. The experimental results show that Pota can save 15% average task completion time while increasing 20% task profit.