BALANCING LOAD ACROSS CACHE SERVERS IN A DISTRIBUTED DATA STORE

摘要

A technology for balancing computing resource load across cache servers in a distributed data store is disclosed. The technology can monitor computing resource load on each cache server to increase or decrease an assigned weight of the cache server. The technology can use two hash functions to reallocate, based on the change in the assigned weight, a portion of the key space from one cache server to another. The first hash function can be a consistent hash function that identifies a cache server mapped to an entity identifier. The second hash function can be employed to determine a hash weight of the cache server. The hash weight of the cache server can then be evaluated against the assigned weight to determine whether the entity identifier should remain mapped to the same cache server or should be reevaluated for allocation to a different cache server.

主权项

1. A method comprising:
distributing a set of entity identifiers among cache servers in a cluster of cache servers, each cache server having associated therewith a subset of the set of entity identifiers; monitoring an operational parameter associated with each cache server in the cluster, the operational parameter providing a measure of a computing resource load on the cache server; adjusting, based on a value of the operational parameter, a weight assigned to a first cache server among the cache servers to balance the computing resource load across the cache servers; and re-distributing, based on the adjusted weight of the first cache server, a portion of the subset of the set of entity identifiers among the first cache server and one or more other cache servers in the cluster to balance the computing resource load across the cache servers.