Caching pages in compute replicas based on aggregate usage of pages across the compute replicas

A system comprising:

• a processor;
• and
• a memory device comprising program code structured to cause the processor to:
• maintain, at a first compute replica of a set of compute replicas, page statistics associated with requests to read or write pages, the requests handled by the first compute replica;
• provide, to a first page server that serves a first page, a first subset of the page statistics corresponding to pages served by the first page server comprising page statistics associated with the first page and an identifier of the first page;
• receive, from the first page server, a first set of aggregate page statistics based on page statistics received from a plurality of compute replicas of the set of compute replicas, the first set of aggregate page statistics comprising, for a second page served by the first page server, an identifier of the second page and an aggregate page access frequency of the second page by the set of compute replicas;
• determine, based on the first set of aggregate page statistics, that the second page is missing from a cache of the first compute replica;
• and
• obtain, from the first page server, the second page for caching at the first compute replica.

A computer-readable storage medium comprising executable instructions that, when executed by a processor, cause the processor to:

• maintain, at a first compute replica of a set of compute replicas, page statistics associated with requests to read or write pages, the requests handled by the first compute replica;
• provide, to a first page server that serves a first page, a first subset of the page statistics corresponding to pages served by the first page server comprising page statistics associated with the first page and an identifier of the first page;
• receive, from the first page server, a first set of aggregate page statistics based on page statistics received from a plurality of compute replicas of the set of compute replicas, the first set of aggregate page statistics comprising, for a second page served by the first page server, an identifier of the second page and an aggregate page access frequency of the second page by the set of compute replicas;
• determine, based on the first set of aggregate page statistics, that the second page is missing from a cache of the first compute replica;
• and
• obtain, from the first page server, the second page for caching at the first compute replica.