The Datawatch Panopticon Server edition supports three levels of caching:
Queryable cache
Recordset Cache
Subscription Cache
All of which are optional. If caching is specifically not desired, data requests can always be forwarded to the underlying data repository.
The Subscription cache describes the cache used for streaming subscriptions. This cache is used to ensure subscriptions are not duplicated by the server, and that instead the server manages duplicate end client subscription requests. Subscriptions are started when the server receives a valid client request, and are only terminated on server instance restart.
The recordset cache, is a traditional key value cache and stores query results from prior data requests to underlying data repositories. The query results are keyed against:
Originating workbook
Originating data table
Supplied parameter values
On Demand Queries
And each resultset is cached with a time to live (TTL) defined in the original data table.
The recordset cache does not store data extracts that can then be further queried, it simply stores query results to minimize underlying data repository load, by minimizing query duplication.