Task Parallel Scheduling over Multi-core System

  • Authors:
  • Bo Wang

  • Affiliations:
  • Department of Computer Science and Technology, Tsinghua National Laboratory for Information Science and Technology, Tsinghua University, Beijing, China 100084

  • Venue:
  • CloudCom '09 Proceedings of the 1st International Conference on Cloud Computing
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

Parallel scheduling research based on multi-core system become more and more popular due to its super computing capacity. Scheduling fairness and load balance is the key performance indicator for current scheduling algorithm. The action of scheduler can be modeled as this: accepting the task state graph, task scheduling analyzing and putting the produced task into scheduling queue. Current algorithms involve in the action prediction according to the history record of task scheduling. One disadvantage is that it becomes little efficient when task cost keeps great difference. Our devotion is to rearrange one long task into small subtasks, then form another task state graph and parallel schedule them into task queue. The final experiments show that 20% performance booster has been reached by comparison with the traditional method.