Icecast Streaming Media Server

[Nonapeptide]

I have a Windows Server 2008 R2 machine running IceCast 2.3.3-kh2. I like to watch the network graphs on busy broadcast times and have nearly had a panic attack when, all of a sudden, the outbound network traffic flatlines. This flatline will last for several seconds. Perhaps ten to fifteen. Suddenly the traffic will erupt ten or more times what the steady flow had been. Here is an example that I was able to take a picture of:



I had just flipped desktops to see my remote connection to the Windows server and was barely able to catch that screenshot before the graph slid away. It does not appear that listeners are being dropped, or at least not enough to cause emails to be sent about disruption of service. This graph represents about 200 simultaneous listeners at 48kbps.

Has this behavior ever been documented? Would this be more likely to be an IceCast issue or a Windows Server issue? I've transferred terabytes of information over many hours and monitored the network traffic but haven't been able to see this kind of traffic pattern for anything except IceCast. I recall this behavior was seen while using an icecast.xml file that had only had mount points and the maximum listener connection options modified. Later on I've modified queue-size and burst-size to deal with laggy clients, however the very occasional flatline and then subsequent spike hasn't seemed to change.

I have yet to implement a monitoring system that is fine enough to detect the flatline and spike that are within seconds of eachother. I can't tell how often it's happening. So far I've observed it three times in the course of several months, but I imagine it could be happening once per day and I'd never know it.

[Nonapeptide]

Curiouser and curiouser!!

Tonight I've kept a remote desktop session up to one of my Windows servers that runs IceCast. I walked by and just happened to catch this in progress:



That entire graph represents about 60 seconds from leftmost to rightmost edges. I caught the flatline right at its beginning. In fact, my RDP session halted for a few seconds; about 5. Once again, this represents about 200 simultaneous listeners and 10Mbps of transfer.

I've never seen any stutter in remote desktop connections even while transferring sustained above 100Mbps using other network application layer protocols nor have I seen any drops in connectivity using any other network application.