- Download 93
- File Size 1.86 MB
- File Count 1
- Create Date December 4, 2017
- Last Updated December 4, 2017
4. Measuring Quality is One Thing, Trusting Our Measurements is Another
Presented by Russ Penar, Principal Service Engineer, Microsoft.
Measuring voice quality and mapping that to a user’s experience is hard; just getting consensus on what metrics to use and the corresponding thresholds can end friendships. Once you’re ready to measure, and away from the homogeny of a laboratory or single vendor ecosystem, you start finding false positives and unexpected circumstances making the quality measurements non-actionable for support teams and customers. Should you have known this was coming?
This presentation provides cases studies from a journey towards getting telemetry ‘right’ for VoIP services when end to end path traverses more than one service/operator network, RTP/RTCP aren’t uniformly handled across providers, media testing/scoring system hiccups are common (e.g. POLQA reference audio with silence periods traversing networks with comfort noise enabled), proprietary metrics are in play, and so on.
The presentation also surfaces debate points on how to ‘best’ monitor service (media) quality; e.g. Transcoding at the edge cleans up metrics with more granular control and reduced surface area, however it also degrades the end user experience by introducing latency.
The content is meant to help others plan for these pitfalls and generate conversation on some of the debate points surfaced.