Hotel IPTV with IGMP snooping problem

[jehanke]

In a hotel IPTV environment we have video servers and setop boxes as streaming clients. The network is build with catalyst 4506 (central switch) and catalyst 3560 (edge switches). All components are igmpv2 speaker, on the switches is igmp snooping enabled. Between the setop-boxes and the switches are broadband cablemodems connected. Up to 5 setop boxes share a switchport. Some times, setop boxes becomes disconnected from network and can?t send a leave for the actual stream. This streams stay for long time and consumes bandwith. Video servers send periodic igmp general queries and setop boxes answers with membership reports.

Is there a change to get the edge switches listen to general queries and memebership reports to find those "ghost streams" and clear them after the maximum response timeout?

[cisco_lad2004]

I dont understand why STBs report when queried ?

but in any case, try fine tunning ur arp time out and also IGMP query intervals.

HTH

Sam

[jehanke]

Hi Sam,

thanks for your replay.

I dont understand the relationship with arp. There is no multicast routing in-between. STBs replay to general queries (send from video servers) and group specific queries (send from switches in response to leaves). The servers send general queries round about every 10s. The response timeout is also 10s. But switches dont recognize this and forward streams to ports where no membership report is received for more than 10s.

[cisco_lad2004]

if arp times out faster, the switch will know there are no STB connected or powered up and a specific port and therefore nothing should be streamed.

[jehanke]

Hi Sam,

the switches are at layer 2 only. They did not know the ip addresses of STBs. They only know the mac address of connected STBs. But this has nothing to with the multicast stream because this stream will be forwarded to the layer 2 multicast group address.

regards

Jens

[cisco_lad2004]

Hi Jens

STB have IP addresses, if they are L3 terminated on 4500 then it is teh place where u should try to fine tune ur arp time out. I am not saying it will fix ur issue for sure, but its worth a shot.

IGMP query interval should however be the 1st parameter to fine tune.

HTH

Sam