I have a video camera (5-10 mb/s) l2 service thats reported loss over a large private microwave circuit. (Along with other services) with a mix of routers. The core is comprised of ccr1009s, and the edge routers are a mix of rb1100ah4 and 3011s. I thought I had a switching issue / layer1 on the camera end, but for a trial, I change the l2 service from VPLS to EOIP, and video loss reported from the NVR dropped to nothing for over a the last two weeks. (The EOIP circuit is over MPLS as well.) Previously, if I reset the loss counters, they would start to increment within a few hours. The routers themselves are not reporting packet loss.
OSPF/MPLS backhaul, all running 6.47.7 and 6.47.8. Only providing l2 extensions for customers. CPU utilization does not ever exceed 1-2% overall, and 5-10% on any one core, on any devices, and its a private backhaul network, so no fw filter rules, just mangle prioritization of ospf/ldp on output. There arent even any queue’s running because its limited to one device.
Any suggestions on why I am finding things to work this way? It seems everything Ive read says vpls should be better, but eoip seems to have less loss?
In this situation, the NVR monitors/reports/records video loss. Its somewhat forgiving, so if its reporting loss, I believe its truly worse than it is.
Packet sniffer on the mpls interfaces shows the control word being used on every packet, which I have accounted for in mtu’s I believe.
I have other vpls circuits with voice traffic on them for a customer which has performance monitoring built into their routers, and switching to EOIP has significantly reduced their complaints for lost packets. Their routers report in 15 minute intervals, and when they complain its ~33 packets lost over a 15 minute interval. As these are continuous voice traffic, they are small in size, but continuous.
I have not tried cisco-style vpls, but I cant imagine that could make a difference. Though I didnt think that EOIP would work better either.
What I would suggest is that, if you are concerned about loss, add a test IP onto both ends of the VPLS tunnel and do a continuous ping from one of the two routers to the test IP on the other side. That way you can see if there is any loss across the tunnel itself - external monitoring could be picking up loss from some other source.