Journal of Beijing University of Posts and Telecommunications

  • EI核心期刊

JOURNAL OF BEIJING UNIVERSITY OF POSTS AND TELECOM ›› 2014, Vol. 37 ›› Issue (5): 20-25.doi: 10.13190/j.jbupt.2014.05.005

• Papers • Previous Articles     Next Articles

Load Balancing Strategy Based on CMM Model in HDFS

LU Mei-lian1, ZHU Liang-liang2   

  1. 1. State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing 100876, China;
    2. Science and Technology Development Center, Hua Xia Bank Limited Company, Beijing 100005, China
  • Received:2013-12-18 Online:2014-10-28 Published:2014-10-28

Abstract:

In order to balance the load of Hadoop distributed file system with lower load migrating cost and data transmission cost, a certainty multi-stage and multi-object (CMM) decision model was proposed. The model is a directed acyclic graph built on decision nodes, which adopts the remaining load capacities of CPU, memory and disk as decision preconditions, and also adopts load balancing effect, load migrating cost and data transmission cost as decision targets. By CMM model, the best balancing plan is determined by selecting results of multiple decision stages and computing the plan usage. Simulations show that the CMM based strategy can achieve better load balancing effect, load migrating cost and data transmission cost.

Key words: distributed file system, load balancing, decision model, remaining load capacity

CLC Number: