Multicast Spillover

[NielsenDBT]

I am a newbie to network switch configuration and my group recently bought an SG 300-28 Gigabit Managed Switch.

It appears by default to spill over (forward) all multicast traffic to all ports.

I want to reconfigure it to only forward multicast traffic to the ports that have joined that specific multicast.

I currently have one connection to our business LAN (10.45.x.x) for configuration on g24.

I have a PC (192.168.1.153) connected to g1 that is streaming out 2 multicast feeds (239.255.5.5:5005 and 239.255.6.6:5006)

I have another PC (192.168.1.159) connected to g13 that can receive 1, 2 or niether multicast feed(s). 

I see that there is a multicast menu in the user GUI, but I'm not sure what items I should be modifying.

Any help would be appreciated.

Doug   

[Tom Watts]

Hello Douglas,

On the switch, if you navigate to multicast FORWARD ALL, what this does is forward all multicast packets if you do not have a bridge filtering enabled on the PROPERTIES page.

If you have the bridge filtering enabled then you can specify the port for a vlan or LAG to direct the traffic. You may choose the option static to join the multicast stream, forbidden to not join or none to not be a forward all port.

Since you're using 2 specific multicast stream, the unregistered multicast shouldn't be an issue, however, you can choose to set the port configuration to FILTERING if you're unsure there is any unregistered multicast (traffic not defined by the address) and this will reject all packets that are not a part of the stream.

[NielsenDBT]

Thanks Thomas,

Perhaps I should clarify that the multicast traffic is successfully getting to the destination PC.  That is not the problem.

What I am trying to do is:

1.  Prevent the multicast traffic from cluttering up our business LAN (the connection on g24)

2.  Save bandwidth on each connection by only sending multicasts to those ports that request to join the multicast.

Although my test is using only 2 multicasts, once I have the switch configured I will put it into a system that has dozens (approaching 100) multicast streams on it.  The multicasts will quickly use up all bandwidth on all ports if they are being forwarded to all ports.

So I read through the Administration guide and:

1. Under Multicast->Properties I enabled "Bridge Multicast Filtering Status"

2. Under Multicast->IGMP snooping I enabled "IGMP Snooping Status" and enabled IGMP snooping on all VLANs using IGMP v3.

3. Under Multicast->Forward All I left all ports in all VLANs set to "None"

4. Under Multicast->Unregistered Multicast I set g1-g24 to "filtering" and left g25-g28 as "Forwarding"

In order to test whether it was working as expected, I went to Status and Statistics->Interface, chose g1 (the multicast source), 15 second refresh rate, and Clear all Interface Counters, then I watched how the Multicast counters incremented each 15 seconds.

I did the same with g13 (the multicast receiver) under various conditions (receiving only 1 multicast, both multicasts, no multicasts, PC turned off, ethernet cable unplugged)

The Multicast TX counter increments about 14k-15k each 15 seconds when both multicasts are received, about 12k each 15 seconds when one multicast is received or no multicasts are received (like when the PC is turned off but the ethernet cable is still plugged in).  It doesn't increment when the ethernet cable is unplugged from the receiving PC.

So I'm not sure if these counters are a good metric, but I would expect the TX count to not increment if the PC is turned off (no multicasts are joined), and I would expect the count difference to be greater when 1 versus 2 multicasts are joined.

Do you have a better way to verify whether the switch is behaving as expected?

Doug

[NielsenDBT]

UPDATE: After repeating some of my tests I am getting results more in line with what I would expect.  I'm not sure what the difference would be except that maybe the "group leave" messages were not getting through or acted upon before.

The 239.255.5.5:5005 multicast seems to take about 3.5k packets/15 seconds.

The 239.255.6.6:5006 multicast seems to take about 8k packets/15 seconds.

I can join one, both or none on the receiving PC and get expected results from the multicast packets/15 seconds rate.