Telemetry
SocketMeister 11.1.0 adds lightweight, thread-safe runtime telemetry for both client and server instances.
Goals:
- Diagnose bottlenecks and support capacity planning.
- Keep overhead low: atomic counters, no hot-path allocations, periodic aggregation.
How it works:
- Atomic counters on send/receive and connection events.
- A background sampler (default 5s) computes rolling averages using an EWMA (~15s window).
- Compression efficiency is observed by comparing compressed and uncompressed body lengths where those lengths already exist in the pipeline.
Metrics (selection):
- CurrentConnections, MaxConnections
- ProcessUptimeSeconds, SessionUptimeSeconds
- TotalMessages (sent + received), TotalFailures (send failures)
- AvgMessageThroughput (messages/sec)
- AvgBitrate (bits/sec; decimal units)
- CompressionRatio and CompressionSavingsBytes
- Reconnects, ProtocolErrors
Defaults & configuration:
- Enabled by default; disable per instance with
TelemetryEnabled. - Aggregation interval:
TelemetryUpdateIntervalSeconds(1–10s, default 5s). - Telemetry does not change the wire protocol and adds no dependencies.
Access patterns:
- For a live view, read
SocketClient.TelemetryorSocketServer.Telemetry. - For a consistent, point-in-time view, call
GetSnapshot()on the client/server.
FAQ:
- Does telemetry change the wire protocol? No.
- What’s the overhead? Typically well under 1% CPU with zero per-message allocations.
- Can I disable telemetry? Yes, per instance via
TelemetryEnabled. - Are latency percentiles available? Not in 11.1.0; intentionally omitted to keep overhead minimal.