IPTV design - switch cascading

[SHABEEB KUNHIPOCKER]

Hello all,

            I have a customer who needs IPTV to be implemented in his network. He needs layer2 multicasting. But we got a design recommendation from IPTV people ( my responsibility is justto configure the switches for IPTV) that we cannot interconnect the edge switches.

my network is designed as shown in the attachment

The IPTV people say that we should connect the edge switches together as shown in red line in the diagram. Please advise on this whether we can connect the edge switches together or not in my design.

Thanks in advance.

Shabeeb

[Jon Marshall]

It's not really clear what you mean. Why do you need to interconnect the access switches. You would only need to do this if the uplinks from the edge switches were routed connections but i suspect they are not.

Interconnecting the edge switches is not really best practice if the uplinks are L2 to the core switches. Can you perhaps explain what you think you get by connecting the edge switches ?

Jon

[SHABEEB KUNHIPOCKER]

Hi Jon,

           Thanks for the quick reply. Am interconnecting the edge switches together because they are in a single rack and we want to have redundancy to the core network. Lets say if the link from sw1 to core 1 fails still Sw1 can connect to the core network through Sw2. This is what i want to achieve and thus we are interconnecting the edge switches.

[Jon Marshall]

Ahhh yes, my apologies. I only skim read the diagram. I failed to notice there was only one uplink from each switch to the core switches. My fault.

A more common approach is to connect each edge switch to both core switches. There is a good reason for this and it is to do with squares vs triangles.

If you look at your diagram you have a square. The problem with a square is this -

lets say that for a particular vlan eg vlan 10 edge1s link to core 1 is forwarding. The link between core1 and core2 is also forwarding as is the link between edge1 and edge2. So STP blocks the link between edge2 and core2 to prevent a loop. Now if the link between edge1 and core1 fails edge1 immediately knows about this because it is direct link. But edge2 doesn't because the link is an indirect link from edge2's perspective. So you are now reliant on timers etc. before edge2 realises that the link has failed and brings up it's previously blocked link to core2.

Compare this with a triangle ie. each edge switch is connected to both core switches but there is no connection between edge1 and edge2. So, again from edge1 perspective. It now has 2 links to the core switches. edge1 to core1 is forwarding. The core interconnect is forwarding. edge1 to core2 is blocking. If edge1 to core1 fails then edge1 can immediately transition it's other link to forwarding because it knows the link has failed ie. there are no indirect links.

So that is why a triangle design is better. Having said all that if you do not have the spare fibers to run from the edge switches to the core switches then yes, the only way to provide redundancy is to interconnect your edge switches.

Edit - note with IPTV the less of a delay in failing over the better but as i say it depends on your available fiber runs.

Jon