中国邮电高校学报(英文) ›› 2021, Vol. 28 ›› Issue (5): 91-101.doi: 10.19682/j.cnki.1005-8885.2021.0025

• Networks • 上一篇    下一篇

Dynamic load balancing algorithm for distributed system

崔岩松; 白春雨;    
  

  1. 北京邮电大学
  • 收稿日期:2020-11-04 接受日期:2021-02-22 出版日期:2021-10-31 发布日期:2021-11-09

Dynamic load balancing algorithm for distributed system

崔岩松; 白春雨;   #br#   

  1. School of Electronic Engineering, Beijing University of Posts and Telecommunications, Beijing 100876, China
  • Received:2020-11-04 Accepted:2021-02-22 Online:2021-10-31 Published:2021-11-09

摘要:

In distributed systems, it is important to adjust load distribution dynamically based on server performance and load information. Meanwhile, gray release and rapid expansion are the basic requirements to ensure reliability and stability for systems with short version iteration cycles. The traditional Hash algorithm performs poorly in gray release, rapid expansion, and load distribution. To solve these problems, a novel Hash-based dynamic mapping (HDM) load balancing algorithm was proposed. On the one hand, this algorithm can adjust the load distribution dynamically based on server performance and load information. On the other hand, it implements gray release by controlling the ratio of requests assigned to the changed nodes. Additionally, HDM has a higher expansion efficiency. Experiments show that the HDM distributes the load more reasonably, provides a more stable gray release ratio, and has a higher expansion efficiency.

关键词: load balancing, gray release, rapid expansion, dynamic mapping, Hash-based dynamic mapping (HDM)

Abstract:

In distributed systems, it is important to adjust load distribution dynamically based on server performance and load information. Meanwhile, gray release and rapid expansion are the basic requirements to ensure reliability and stability for systems with short version iteration cycles. The traditional Hash algorithm performs poorly in gray release, rapid expansion, and load distribution. To solve these problems, a novel Hash-based dynamic mapping (HDM) load balancing algorithm was proposed. On the one hand, this algorithm can adjust the load distribution dynamically based on server performance and load information. On the other hand, it implements gray release by controlling the ratio of requests assigned to the changed nodes. Additionally, HDM has a higher expansion efficiency. Experiments show that the HDM distributes the load more reasonably, provides a more stable gray release ratio, and has a higher expansion efficiency.


Key words: load balancing, gray release, rapid expansion, dynamic mapping, Hash-based dynamic mapping (HDM)