Network metrics

[Wiezzel]

General remarks

• All metrics are sent via libp2p pubsub and stored by the router.
• Enabling metrics is opt-in for the client.

Worker metrics

1. Every 10 seconds:
   • Available datasets/ranges,
   • …
2. For every received query:
   • Submitting client’s peer ID,
   • Query ID,
   • Query itself,
   • Dataset ID,
   • Size and hash of input data,
   • Execution result (OK or error),
   • Execution time,
   • Size and hash of output data.

Client metrics

For every submitted query:

• Query ID,
• Query itself,
• Dataset ID,
• Worker’s peer ID
• Execution result (OK, error, timeout),
• Execution time (incl. result download time),
• Size and hash of output data.

[dzhelezov]

For the query data, we also need the size of the scanned data, not just the size of the result.