Networking VLAN traffic query

Banzai_Joe

Banzai_Joe

Banzai_Joe

Soldato
Joined
30 Sep 2006
Posts
5,278
Location
Midlands, UK
Hi all,
so our network of 16 Aruba switches is mostly a physical ring topology (so bi-directional). Some switches do have uplinks to other switches within the ring, so kind of giving a mesh topology for those portions of the network.
I recently had an issue whereby a couple of devices on a specific VLAN weren't grabbing an IP address from the DHCP server on our UTM.
I eventually tracked it down to one particular switch not having the relevant ports tagged for that particular VLAN.
Here's my query; why would the traffic NOT go the other way to its destination? once it hit a brick wall in one direction, why not go the other way round to get to the UTM?

That's it really?

any advice would be appreciated, thanks.
 
To turn the question around, why should the traffic go the other way? I assume you're using STP or RSTP? It'll be using physical link state to determine the root bridge so that has no idea if your VLAN tagging is correct or not.

If you wanted something a bit more clever you'd need to look at something like IP SLA with BGP.
 
ChrisD. said:
Most likely spanning tree, you shouldn't be deliberately creating a loop with switches. Why are they configured in a ring?
Click to expand...
It's a football stadium, so imagine looking from a birdseye view, there are switches dotted around the perimeter. Its a fibre backbone but physically its more of a ring topology. the mesh topology is present where the main comms room is, purely because there are more switches within a smaller area so its easier to mesh them up if you get my drift.
STP is on by default.
 
Last edited:
I don't admit to know much about networks in stadiums but similar principles to regular buildings should apply. Core, distribution, and access is still prevalent here as it's not a datacenter topology.
 
ChrisD. said:
I don't admit to know much about networks in stadiums but similar principles to regular buildings should apply. Core, distribution, and access is still prevalent here as it's not a datacenter topology.
Click to expand...
Lol, i really ought to finish my CCNA, but i've lost half of the info i've learned from simply not having to apply it.....until now.
 
The reason why you saw the specific behaviour that you had is likely because you are using STP instead of PVST, but I wouldn't bother enabling that. Rather than chaining switches through each other, ideally you'd have a core and then all your access switches connected back to this core - you can use fibre and have one link go clockwise around the stadium and the other go anticlockwise if you want. Then the two fibre uplinks on each access switch are in a link aggregate and your core is either a stack or you're using switches that can do MLAG.

Having to diagnose every possible switch on the path is making things more complicated than they need to be. You also massively increase the bandwidth available if each switch has 20Gbit (for example) back to the core, rather than everything sitting on a single 10Gb link as it works it way back to the comms room.
 
Last edited:
Network map

I inherited this, but have made changes as budget allows.

@Caged - most of our switches are 8-port with 2 x 1gb SFP's, would you mind expanding on what you said about the 2nd fibre circuit going in the opposite direction please?
I guess i assumed that a single fibre ring would allow traffic to go in both directions.
 
Last edited:
Caged said:
The reason why you saw the specific behaviour that you had is likely because you are using STP instead of PVST, but I wouldn't bother enabling that. Rather than chaining switches through each other, ideally you'd have a core and then all your access switches connected back to this core - you can use fibre and have one link go clockwise around the stadium and the other go anticlockwise if you want. Then the two fibre uplinks on each access switch are in a link aggregate and your core is either a stack or you're using switches that can do MLAG.

Having to diagnose every possible switch on the path is making things more complicated than they need to be. You also massively increase the bandwidth available if each switch has 20Gbit (for example) back to the core, rather than everything sitting on a single 10Gb link as it works it way back to the comms room.
Click to expand...
Budget just isnt there i'm afraid. That said, cabs B, F-J are at the tea bars and are only used at home matches, so traffic throughput will never really be an issue.
 
Caged said:
The reason why you saw the specific behaviour that you had is likely because you are using STP instead of PVST, but I wouldn't bother enabling that. Rather than chaining switches through each other, ideally you'd have a core and then all your access switches connected back to this core - you can use fibre and have one link go clockwise around the stadium and the other go anticlockwise if you want. Then the two fibre uplinks on each access switch are in a link aggregate and your core is either a stack or you're using switches that can do MLAG.

Having to diagnose every possible switch on the path is making things more complicated than they need to be. You also massively increase the bandwidth available if each switch has 20Gbit (for example) back to the core, rather than everything sitting on a single 10Gb link as it works it way back to the comms room.
Click to expand...
Hi mate, been thinking on this. So currently I have a single OM4 cable looping in and out of each switch (daisy-chained). Can i use the same cable but a separate fibre pair to achieve the bi-directional circuit as you suggested but without the need for a 2nd cable.
I guess we'd still have a SPoF as we only have one cable (and rats do live here) but would it alleviate the issues if one switch were down as the traffic would simply try going the other way?
Advice appreciated (as always).
 
You must log in or register to reply here.
Back
Top Bottom