Backup Interface / HSRP? / EIGRP - anything else?!?

wlegge · ‎03-18-2004

Hi all

I have been racking my brain for a couple of days now about a network plan that we are thinking of - currently I only have one solution and wonder if there is a better one!

See my attached .jpg for the net diagram...

What I would like to be able to do is use 200mb between our sites (to have full resiliance) - this is what I am not sure about.

What I can do (and have tested) is to put all of the interfaces in the same VLAN and then use 'backup interface's to control the loops, and to provide us with some kind of fault-tolerance.

Before anyone suggests why I am not using spanning-tree to look after the loops etc - we have tried that although the switch puts both ports to forward. I think this is because the telco has provided us with a VLAN already (our termination from the telco is a WWP LE46 box with 100BaseT port for us).

Anyway, to this end we cannot have both ports on the same switch in the same VLAN (or if they are, they cannot be both active at the same time).

Currently what we have is two VLANs (one for the ports connected to the Telco Core A and one for the ports connected to the Telco Core B). These are obviously numbered at the moment and we have full connectivity - however what we do not have is the ability to use one 'path' over the network.

Any ideas? I have thought about Etherchannel, HSRP etc but cannot think of the right solution.

Thanks

Will.

PS - Damn stuff comes along too quick - I became CCNA a month ago and am on 2/4 CCNP books now, must read faster!!!

konigl · ‎03-18-2004

Let me see if I understand your diagram correctly. You have three sites -- A, B, and C -- and a single telco core location. The telco core location has two core switches, I'll call them CS1 and CS2; they connect to each other via a 1-Gbps link. Each site has two switches: one that connects to CS1 over a 100-Mbps link, and the other that connects to CS2 over a separate 100-Mbps link. Everybody connected to CS1 is on VLAN 1, and everybody connected to CS2 is on VLAN 2. (The numbers may not be the actual VLAN ID numbers; but I don't want to confuse them with your site designations so I'm using the CS numbers I defined above.)

What kind of switches are at the sites?

What protocols are you using? (TCP/IP? IPX/SPX? Appletalk? Other?)

Are you trying to connect the two switches at each site to each other? Or are you looking to connect your sites A, B, and C to each other and independent of the telco core switches? Or are you looking to double the bandwidth of the connection from each site switch back to its telco core switch?

Are the telco core switches acting as default gateways or routers for your VLANs?

Do you know if the 1-Gbps connection between the two core switches is Layer-3-only, or is it a Layer 2 VLAN trunk, or both?

What kind of switches are the telco core switches?

I need more information before I can recommend EtherChannel vs. STP vs. HSRP vs. a dynamic routing protocol with load-balancing across equal-cost paths vs. something else. Also, need to understand exactly what you're wanting to accomplish.

For example, both switch ports going to forwarding could be because your switch became the root bridge for your VLAN's spanning tree. Or it could be because some other switch is the root, but there is no Ethernet loop that needs blocking on your switch. Or it could be because there is an Ethernet loop, but it's blocked on a different switch than the one you were looking at. Or it could be because you bridged VLAN 1 and VLAN 2, creating one giant extended broadcast domain out of two smaller ones. I would need to know more about the network and your intentions in order to rule out some of these possibilities.

wlegge · ‎03-19-2004

Hi

Thanks for the response.

Firstly - as you have seen, this is bit complex to explain!! What we are being actually given by the telco is a totally flat network with no VLANs put in place by them what so ever.

All ports (A,B,C) on both CS1 and CS2 can all see each-other. The CS1 and CS2 are in two physically different locations (about 5km apart) to give us resiliance.

What we are trying to achieve is that sites A,B and C can all see each other (they have their own local VLANS and subnets). If we had just CS1 - that would be easy - job done.

But! What we were thinking was that the connections to the CS2 will give us some resiliance if the CS1 has a fault, or if a link between A,B or C to CS1 has a fault.

Ideally, if we can use both the link from A,B and C to CS1 and CS2 concurrently, (doubling bandwidth) that would be excellent.

We have one switch at each of our sites - A has a Cat. 4000 w/various blades, B has a Cat 3550-24-EMI and C has a Cat 3550-12T (also EMI). We only need IP traffic.

The telcos core switches are WorldWidePackets LightningEdge 327's and the 'local' telco equipment in our sites are LightningEdge47's.

BTW - on the STP side, when we were trying this we only were using one VLAN across the whole solution, and we had site B plugged into CS1 only and then site A plugged into CS1 - that works fine, but when you plug in site A to CS2, it puts that port into forward too (loop). Obviously portfast etc is not enabled. Do you think the telco's kit would interfere with the spanning-tree?

Thanks

Will.