The third issue of our popular reference guide for world wide WebRTC performance metrics is available as a free download. Monthly newsletter subscribers will get the report to their inboxes.
As WebRTC matures, we begin to see exciting developments in the field. To wider circulate these findings, we began to publish a regular WebRTC Metrics Report series that presents key data and insights on the sessions monitored by callstats.io.
Use cases of WebRTC are diversifying, from collaboration to two-sided marketplaces to contact centers. During the first half of 2017, our install-base continued to grow rapidly, with an average 20% month on month growth, and the overall monitored traffic grew by 5 times. Consequently, as callstats.io’s Monthly Active Users and data volumes grow, we are observing an increasing demand for more nuanced metrics.
For example, companies building new WebRTC products want to know their peer-to-peer topologies’ and bridge-based topologies’ performance. Existing services are pursuing hybrid models: First, using direct peer-to-peer opportunistically and when there are performance issues (e.g., connectivity issues or more users) the clients switch to using a centralized bridge to deliver media to one or more participants.
Highlights from the report
- Chrome still dominates the WebRTC browser market. Likewise, the operating system distributions remain the same, with Windows leading.
- 40% of the connectivity checks finish in 100ms.
- 80% of the media sessions render media in less than 1s after starting the call.
- We have an even split between bridge and peer-to-peer traffic. 50% for each network topology.
- Average Round-trip time is lower for peer-to-peer sessions than the sessions using the bridge.
- Median average RTT for peer-to-peer sessions is 70ms and it is 140ms for bridge sessions.
- 95-percentile fractional loss is lowest for the bridge followed by peer-to-peer, and highest for TURN sessions.
- About 90% of the sessions on the bridge, 84% on the peer-to-peer, and 78% on the TURN have zero 95%-ile losses.
WebRTC insights for better communication
By putting our metrics into a report we hope to serve the community to build bigger and better WebRTC applications. We expect the report be useful for engineers and product managers alike in their efforts to design better real time communication systems. As the WebRTC community prepares for the growth ahead, these insights will ideally help circumvent potential bottlenecks.