Understanding persistence stores

Persistence stores are a special category of publisher that can be used to publish updates from a persistent backing data store, such as, a database. A persistence store is always able to provide an immediate integrity update for a subscribed topic by retrieving the data from the backing data store. A persistence store does not need to wait for the data to change.

Persistence stores connect directly to the PubSub Engine running on the same computer, using the Local Delivery Layer (LDL) for communications. When a persistence store starts, it registers itself with the local PubSub Engine, specifying the name of the topic space for which the persistence store is responsible. The persistence store becomes the local publisher for all topics under the registered topic space. The PubSub Engine works in tandem with the persistence store, forwarding the appropriate subscription requests and relaying publications back to the subscribers.

There should only be one publisher at a time responsible for the persistence store’s topic space. If the persistence store belongs to a redundant service, it should only register the topic space when the service is Hot and should de-register when the service goes into standby.

If the PubSub Engine has any existing subscriptions under the registered topic space or receives any new subscription requests for it, they are handed off immediately to the persistence store. The persistence store is then responsible for tracking the subscriptions and publishing updates for them.

When a persistence store disconnects from the local PubSub Engine, or explicitly de-registers itself as a persistence store (for example, when the service goes into standby), then the local PubSub Engine resumes responsibility for subscriptions that formerly registered the root topic.