Described herein is a novel technique for implementing a policy-based caching engine in a storage system cluster (cluster) to automatically implement volume caching at select nodes of the cluster in optimizing cluster performance. The novel caching engine may be implemented in a management console of the cluster storing policy information related to volume caching operations requested by the caching engine. Policy information may include node or cluster attributes, operational events, and a pre-defined cached volume configuration of import to the user. Based on the stored policy information, the caching engine may determine a set of nodes on which to initiate a volume caching operation and generate a request for such operation to be performed on a particular node. Only select nodes in the cluster may thus implement volume caching to thereby conserve processing resources for more critical operations at other select nodes in the cluster.