Performance Metrics

[zachkinstner]

Rexster provides powerful monitoring capabilities, which can be sent to monitoring tools like Graphite. RexConnect could also collect/send various metrics:

• information about TCP connections and throughput
• query response times
• heartbeat response times (and success/failure events)
• request/response sizes
• JSON de/serialization times

This could help identify potential bottlenecks related to the RexConnect TCP layer.

RexConnect could provide various configuration options for these metrics. Perhaps it could utilize the functionality already built into Rexster.

[zachkinstner]

When running in embedded mode (see #23), RexConnect could also collect and broadcast metrics from Cassandra. This would simplify deployments, as it could remove the need to install/configure additional software to do this task. Maybe RexConnect could embed a utility already built to do this -- based on this discussion, maybe CollectD is the right tool. More investigation needed.

[zachkinstner]

Here are some notes from DataStax about sending Cassandra metrics to Graphite.