Dynamic allocation consistent hashing algorithm based on virtual node ranking
Zeyun Sun, Zhenghe Liang, Yingying Wu · 2021
In the Internet environment, e-commerce and other information enterprises are developing rapidly, and data information is growing exponentnently. Because of the fixed structure, poor scalability, uneven load, low resource utilization and other problems, the traditional database cluster can not bear the load pressure caused by sudden data surge, which may lead to cluster collapse and business paralysis. In order to solve the problem of load imbalances caused by the random calculation of hash function and the difference performance of each node, this thesis proposes a dynamic allocation algorithm based on virtual node ranking (VNRCH) of consistent hash. According to the CPU, memory and bandwidth of the current node, the comprehensive load of the node is calculated, and then the number of corresponding virtual nodes is calculated according to the load. The dynamic distribution of virtual nodes combined with the improved allocation with virtual node ranking can not only promote the cluster load balance, but also avoid the excessive access of nodes. Experiments show that the algorithm improves the performance of load balancing, response time and fault tolerance, and enhances the stability and robustness of the cluster.