Packet loss snapshots as seen from NCSA, 7/8/1999

It's a moderate annoyance for video, but a severe problem for audio -- the dropouts are much too long for rat's buffering to carry over. When this is happening, it's not very practical to carry on a conversation.

This isn't only an issue today -- it's been true most of the time, during the access grid mbone tests I've seen over the last week or two, and probably is much longer-lived than that.

It's not specific to the access-grid sources either; I see a similar loss pattern from the NASA Learning Channel.

The periodicity is quite consistent -- there's a loss burst every 5 seconds, typically lasting for 1-2 seconds, during which no RTP packets at all are received, judging from the "Packets" and "Missing" graphs.

I wonder whether the periodicity can be used to diagnose the source of the loss? What happens every 5 seconds?

	Argonne FL Library 140.221.8.53, 7/8 11:15CDT 224.2.177.155	U. Kentucky 140.221.9.169, 7/8 11:05CDT 224.2.177.155	NASA Learning Channel 128.102.84.134, 7/8 11:11CDT 224.2.190.243	More NASA (less severe now) 7/8 11:36CDT
Packets Received		(Didn't grab these)
Packets Lost
Stats

Stuart Levy, slevy@ncsa.uiuc.edu