A Low-overhead Cluster Management Mechanism Based on Node Information Storage of Max-Heap Tree

In order to solve the problems that node information occupies too much storage resources and heartbeat detection mechanism makes system operate under a heavy load in traditional distributed systems, a low-overhead cluster management mechanism based on node information storage of Max-Heap Tree is established. A structure of Max-Heap Tree is applied on nodes in cluster to store node information and an efficient heartbeat detection mechanism is proposed. Besides, node capability parameter is added to node information to evaluate the capability of nodes, so that the failure recovery mechanism can be established by it. The experimental results and analysis show that the mechanism of node information storage based on Max-Heap Tree reduces the occupation of storage resources of the cluster comparing with the traditional storage method. Meanwhile, node and link faults of Master are distinguished in the related heartbeat detection mechanism which reduces the load of cluster significantly when the number of nodes in cluster is over three. Subsequently, the complexity of the algorithm for fault recovery mechanism is greatly reduced.