Skip to main content

Rationale

Without going into deeper details on what concerns Data Orchestrator's DB internals, there are some considerations one should always keep in mind when designing interfaces.

Querying an Interface is fast, querying across Interfaces is painful​

Data Orchestrator's data modeling is designed to optimize queries within a single interface. Querying across interfaces is supported, but might affect performances significantly, especially if done frequently and with complex queries. This is especially true for triggers, as they could be evaluated very frequently.

In general, if you plan on having different mappings which are frequently queried altogether, or dependent on each other for several triggers, you might be better off in having them all in the same Interface.

Aggregation makes a difference​

Aggregation is a powerful feature, which comes with price and benefits. Even though each series has only one timestamp for all values, it is also true that losing granularity for endpoints might cause storage of redundant data if only one of the aggregated mappings change value.

Moreover, in terms of data modeling, Aggregated interfaces imply the creation of a dedicated Cassandra table. Having a lot of aggregated interfaces might end up putting additional pressure on the Cassandra Cluster in terms of memory and overall performance. Your Cluster administrator might (rightfully) choose to limit the amount of installed aggregate interfaces in a Realm, or in the overall Cluster.