2018-07-03 Meeting notes: sync on CORE-598 (Investigate doing less work in RSpace to report metrics)

Resources

• https://rchain.atlassian.net/browse/CORE-598

Discussion

• RSpace team is looking for ways to optimize calculations for throughput (produce and consume)

  • Currently using --diagnostics

  • Going forward idea to more appropriately use Kamon and Prometheus to collect and report metrics

• We have the flag --diagnostics in node

  • Diagnostics was not intended to be used for performance metrics

  • IDEA to remove --diagnostics flag (ticket)

• Clarification - Goal is to move calculation of metrics out of RSpace

• Discussion about using either spans or histograms for measuring events

  • For measuring throughput we need to measure events across time and attached to a point in time

  • Recommendation to use timestamp counters

• Idea if Prometheus does not provide the requisite granularity, consider StatsD

• Requirements

  • Today we need:

    • baseline measure of throughput

  • In future we can iterate to include:

    • Throughput

    • Peak