一种改进的主从节点选举算法用于实现集群负载均衡

A master-slave election algorithm for load balancing clusters

随着数据量的不断增长,分布式搜索引擎的出现满足了大数据量的检索性能.现有的主从节点选举策略往往会导致主从节点分配不均而使得查询性能不均衡.现提出一种搜索引擎集群的主从节点选举策略,能保证正常情况下节点的分布均衡.当出现宕机时能重新切换主节点,保证检索功能有效;当宕机的服务器重新上线时,主从节点分布恢复正常,避免了负载不平衡的缺陷,达到合理利用每台服务器的性能并充分发挥集群的性能优势的目的.

With the increasing amount of data, distributed search engines appeared to meet the need of big data retrieval performance. The existing master-slave node election strategy leaded to the imbalance of query performance due to uneven node distributions. To solve this problem, we proposed a master-slave election strategy based on search engine clusters. When a server was down,a main node was switched to guarantee the effectiveness of the retrieval And when the broken server was back online again, the distribution of the master-slave node also came to normal. Thus, the defect of load unbalance was avoided. At the same time, every server was used reasonably and we could take full advantage of the clusters.