Recently we had a live webcast event go horribly wrong. I'm doing the technical post-mortem, with limited information.

We know that the hardware encoder (a Digital Rapid Touch Stream Web HDI) was unable to send upstream at a sustained reliable high rate. What we don't know is if the encoder's connection was problematic (Zürich), or that of the streaming server (in Frankfurt). Unfortunately, I've got three different vendors all blaming each other (the CDN who runs the server, the on-site ISP and the on-site encoding team.)

In the FMS log files I see a couple of interesting things:

• Zillions of Status Code 408 on play event entries for clients. Adobe's documentation stats that this "Stream stopped because client disconnected". ("Zillions" would be a ratio of 10 events for every individual IP address.)
• Several unpublish / (re)publish events per hour for the encoder

I'd like to know if all those 408s could tell me with authority that the FMS server was starved for bandwidth, or that the encoding signal was starved (and hence the server was disconnecting clients.)

Any clues?