摘要

In this paper, we analyze the conventional dispatching algorithm of LVS (Linux Virtual Service) clusterweighted least-connection scheduling algorithm. This algorithm only considers the number of current connection service node as parameters which is not able to reflect the server load well. In order to describe the situation of the server load better, we propose a novel weight distribution scheduling algorithm to improve the weighted least-connection scheduling algorithm. The novel algorithm uses two parameters including the CPU occupancy rate and memory free rate to describe the load nodes situation. This paper introduces the process of producing the improved algorithm, and analyzes the implement of main modules of the algorithm in detail. The novel weight distribution scheduling algorithm optimized the LVS cluster's load balancing problem, so that the performance and stability of the LVS cluster is largely improved.

  • 出版日期2011

全文