Currently, system metrics include all messages processed by the system, including internally generated messages used to invoke referenced services. The resulting metrics are not completely intuitive (e.g. send in a soap message, but see two or more messages processed in the system metrics). In addition to message counts, total time metrics do not add up either, as the service time does not include the time spent invoking references.
- System message count should only include messages received through gateways.
- Timings for services should be the total time spent processing the message, including time spent in any referenced services.
I think we should only focus on first level reference for the 0.4 release. Longer term, we should look at a hierarchical view. We may also look at other types of aggregation within a service (e.g. by operation).