This is the second set of videos created by the 8x8 University. To follow the analysis on callstats, you can sign up for the demo at the try.callstats.io.Modules 1 and 2, explained the structure of callstats dashboard, and a workflow for investigating call quality issues in a call. The modules in this blog post follow the instructions provided in the video.
Scenario 1: Call Setup Failed, Multiple rejoins (churn)
This video covers the simple scenario of identifying issues with call set up failures or when a user repeatedly drops off from the call.
Totally failed conferences occur when no connections are successfully established between any participant in a conference. Which usually indicates that the participants attempted to join, but were unable to join the conference but due to various types of errors: Media source related, Negotiation related, Firewall/NAT related.
Partially failed conferences occur when connections for some participants fail to establish, meanwhile succeed for others, this is most likely to occur in multi-party (more than 2 participants) conferences, because in a two-party call, if one person fails to join the conference is categorised as totally failed. Below shows the various stages of call set up failure.
Churn is the total number of times a user has rejoined the conference. For example, when there are problems in a particular conference, the user may choose to drop out and rejoin the conference in an attempt to solve the issue. This is considered churn.
Scenario 2: Connection & Call Quality Issues
Connection issues stem from either losing connectivity mid call resulting in a dropped connection or automatically re-establishing connectivity mid-call, where a participant is unable to hear or speak for a few seconds (called disruptions).
Network disruptions can happen when the network changes during the call or if the enduser if far away from the access-point and the network fades in and out or if you are mobile and the cell towers do soft- or hard-handovers. In all cases, the audio/video will disrupt from a few seconds up to 10s of seconds. Typically, if the connectivity does not re-establish in 30-35 seconds, the call will drop.
Media disruptions are identified when the network is connected but the endpoint experiences very high packet loss (50+% loss), or very low throughput (typically, lower than the operating zone for a codec, for example, opus struggles when bitrates are below 15 kbps), high jitter (variations higher than 150ms-400ms for audio might result in jitterbuffer resets) this not rendering audio for several seconds
Over time, you will discover other steps and tricks that will help to isolate issues in your specific work environment. Please share those with us on Twitter or via an email, we look forward to learning from the community!