Multicast routing-realy basic

[Boris Krajnc]

I  start configuring Cisco 2821 router for multicast . First short  description and attached sheme explanation. Let we say  I have small  network with 100 users. One router and Cisco switch 3560. Two VLAN’s,  one for data another for multicast. Data from internet works fine but  now I want to connect multicast servers (or source of more multicast  streams) from another subnet. Router have three interfaces.

I expect there should be no problems with multicast configuration, but unfortunately it is not like I expect. What I did ?

First step:  enable multicast routing

Second step: on both interfaces (Fe 0/1 and Fe 0/2)  - ip pim sparse-mode

Third step: configure switch that users are connected to access port in VLAN 222 (temporary to see if multicast work)

When  I start VLC on computer nothing happend. If I try to connect computer  on same subnet where is source of multicast streams it works fine.

What  I am doing wrong ? Is there anything about routing ? All subnets are  directly connected. RP is not needed if I have one router or ?

Any advice would be very helpfull.

Boris

[Boris Krajnc]

....nobody using multicast ?

[Peter Paluch · Accepted solution]

Hello Boris,

Even if there is just a single router, you should configure a RP to avoid any interdependencies in the IOS implementation. Simply create a loopback on your router and refer to that loopback's address when configuring a RP:

interface Loopback0

ip address 192.0.2.1 255.255.255.255

!

ip pim rp-address 192.0.2.1

In addition, make sure that the VLC is configured to stream the multicast with the TTL set to at least 2. A couple of previous VLC releases was streaming the multicast with TTL set to 1. There was a bug in the GUI that caused the TTL setting to be ignored and not carried over to the streaming settings. Try using the most recent VLC version.

Best regards,

Peter

[Boris Krajnc]

Peter, thank you very much for your answer. I will try tomorrow in my lab and I will report about results.

Best regards

Boris

[Boris Krajnc]

Peter,

Yesss, I did what  you suggested and it works perfect. Very important is also how TTL on VLC is set. By default it was 1 (didn't work) and when I change it on 2 then starts to work. I didn't expect everything will be like I want  and of course I ask three users to trying start the VLC and watch 239.1.1.5:5000 (all users one stream). It also works at all three users without distorted image or any prroblems but (but - why we always need this word ) I watched how much traffic I  see on output interface. If one user watch stream there was approx. 1.5 Mbps if there was two users there was around 2.8 Mbps with three users 4 - 4.5 Mbps. I would like to avoid this and of course if all users watching one stream  they use approx. 1.5 - 2 Mbps.

How to do that ? Some hint would be more than helpfull.

And again Peter,  thank you for your advice from post before.

Best regards

Boris

[Peter Paluch]

Hello Boris,

This is strange. Are you saying that the amount of output traffic grew with the number of recipients? This is definitely not how the multicast should work.

Assuming that there is only a single router between the source and the destination, and all receiving stations are in the same network then regardless of how many receiving stations are in the destination network, the amount of data flowing into that network should remain constant.

How did you actually measure the amount of traffic? Perhaps the methodics of measuring you have used has some deficiences.

I am very glad you got it running!

Best regards,

Peter

[Boris Krajnc]

I totally agree with you but it looks like every user have its own session what of course is not like multicast work. I can not say for 100% ( not yet) that my measure is correct (it was measured on access point on which users was connected). Yes, I upgrade my network a little on part where user is connected to switch (please see attached file). I added wireless connection (access point - station) and there is only L2 between AP and client and I think this should not affect to multicast.

[Peter Paluch]

Boris,

I suggest measuring the flow at the router's Fa0/2 interface or on the switch, but not on the WiFi access point. The WiFi is somewhat more complex technology and I am not entirely sure about the multicast delivery details in these networks.

Best regards,

Peter

[Boris Krajnc]

Hi Peter,

I will research and report about findings on Fa0/2, but I think it is quite normal what I saw. Access point act like a switch and users connect to it. That's mean that I saw flow from all users like if I measure every port on switch and result was cumulative flow...

First I have to see what happened on router.

Thank you for your patience and answers.

Regards

Boris

[Boris Krajnc]

Hi Peter,

I think everything works fine. Regarding to what I can saw on measuring points I can say, that multicast stream from router does not growing with number of users (if they watch same stream). On AP site is like I thought... every user on AP is like he is connected on single interface and I see cumulative flow.

One more question about Join and Leave. Join working fine but after user Leave session I can see stream for a while. How to stop stream immediately?

Regards

Boris

[Peter Paluch]

Hello Boris,

Answering somewhat late - I apologize for that.

One more question about Join and Leave. Join working fine but after user  Leave session I can see stream for a while. How to stop stream  immediately?

In IGMPv2, after a host leaves the group and sends a Leave message, the router will send a group-specific Query message to find out if there is anyone else still subscribed to the group. If no answer is received to a time limit, the router will stop delivering the multicast data to the network. The time delay the router waits for a report to come in is the time you still see the multicast flowing.

There are two ways of limiting this time. The first one is to shorten the time interval during which the router expects a response to its group-specific Query. This time limit can be configured on a per-interface basis using the ip igmp last-member-query-interval command in the range of 100 msec to 25.5 sec. By default, a router sends two group-specific queries in the interval defined by this command, and after the second query, if no response is received within this timeframe, the multicast flow will be stopped. Setting the last-member-query-interval to, say, 500 msec will cause the multicast flow to be stopped after approx. 1second (2x500ms) after the last member leaves the group.

The second way of solving your issue would be to use the ip igmp immediate-leave command. This command will cause the router to stop the multicast flow immediately after receiving a Leave message. However, this command is suitable only for a single-host network. As any single member leaving a multicast group will cause the multicast flow to be stopped, other members (if there are any) will experience interruptions in the multicast service. Therefore, if you have multiple receivers on the same segment, this approach would not be appropriate for you. The ip igmp immediate-leave command requires you to specify an ACL describing a list of groups eligible for immediate leave processing.

Best regards,

Peter

[Boris Krajnc]

Peter,

What can I say. Thank you very much  for your detailed explanation. I will follow your advices in my configuration.

Best regards

Boris