Load balancing method and device

Abstract

The invention discloses a load balancing method and device, and relates to the technical field of cloud computing. The specific implementation scheme is as follows: obtaining load information of eachnode in a cluster, wherein the load information comprises at least one of the following items: a hard disk load, a network load, a CPU load, a memory load and a delay load; calculating a load value ofeach node according to the load information of each node; for each node, diagnosing the node according to the load value of the node to obtain a diagnosis result; conducting processing from the diagnostic result. According to the embodiment, cluster heterogeneous and hot spot data scenes are met, meanwhile, cluster loads have higher fault tolerance, and the loads are more uniform.

Description

Technical field[0001]This application relates to the field of computer technology, specifically to the field of cloud computing technology.Background technique[0002]The distributed transmission load balancing scheme refers to the strategy of how partitions (Partition) are distributed among the nodes (Broker) of the cluster in the distributed transmission system. In actual business, the load of nodes needs to be balanced to improve the availability of the system. At present, the industry generally uses polling, consistent hashing, and dynamic scheduling to solve load balancing problems.[0003]However, the polling method cannot cover the scenarios of node performance differences caused by cluster heterogeneity. In the consistent hashing method, the weight of each Partition in scheduling is the same, and the difference in other indicators of the Partition (such as QPS and traffic) is not considered, and the mapping relationship of the consistent hashing algorithm cannot be interfered an...

AI Technical Summary

Problems solved by technology

[0003] However, the polling method cannot cover the node performance difference scenario caused by cluster heterogeneity

In the consistent hash method, the weights of each Partition in the scheduling are the same, and the differences in other indicators of the Partition (such as QPS and traffic) are not considered, and we cannot intervene and adjust the mapping relationship of the consistent hash algorithm

The dynamic scheduling method does not quantify the migration of Partitions, which may easily lead to over-migration or under-migration. Moreover, the consideration of disk selection is single, and hard disk IO bottlenecks are likely to occur.

Images