DSLw for transparent redundancy?

network · ‎05-14-2002

I'm working on rolling out a redundant lan for a multipop cable co. Currently we're using all Cisco equipment for routers and UBRs, and HP switch gear for ethernet. We've got three pops, A, B, and C. Each pop has at least one UBR. Pop A is special as it has the border router for the network. The pops are wired with two sets of fiber, daisy chained, pop A to B, B to C. Right now we've got one half the fiber lit, and running ethernet over it. In each pop, there are two HP 4000M switches, which are not connected to each other, powering the fiber through gig-e transcievers. This gives us two redundant gig-e switched WANs in each pop running in parrael. As of right now, we're only riding on one of them, and only one is connected to the gateway router in pop A. All the routers/ubrs use OSPF to advertise/share routes amongst themselves and with our primary router which is at annother site, directly connected to the gateway router in pop A.

My first step in adding some redundancy is going to be throwing a crossover cable between the two switches in each pop and using HP switch meshing to keep from having loops. This will allow traffic to flow even if there is a single fiber cut between each pop. (The fiber is all on diverse paths physically so this should normally be the case.)

My second step would be to use the backup interface functionality of IOS to allow each router/ubr to straddle both switches, so should a switch loose power and go link dead, again, connectivity is not interrupted.

Problem is, this setup does me no good if a switch just hangs itself, but maintains link status on the connected ethernet links. The backup interface will never fire up under this scenario. Recovery would require manual intervention which I'm looking to avoid. HSRP looks to be out as afik it isn't designed to switch between interfaces on a single router. DSLw at first glance appears to be my solution, but I'm still digging. Does this seem like a good match, or should I be investigating something else?

If it helps, we have plans for a second diverse uplink to pop A and have no qualms terminating it in a second gateway router. We do need to loadbalance traffic across these two links however when possible. (IE if they are both up they both need to be utilized)

rickwilliams · ‎05-18-2002

Hi GWI,

You might be interested in a new feature coming in Cisco IOS. It's called Gateway Load Balancing Protocol (GLBP). It was recently annouced as part of the Globally Resilient IP launch.

Gateway Load Balancing Protocol (GLBP) is a Cisco innovation that uses available network bandwidth effectively by load sharing packets over all available circuits while protecting the first hop router. Previously, first hop redundancy features only forwarded packets over the backup WAN circuits if the primary router or primary circuit became unavailable. GLBP load balances IP traffic across multiple routers in the Enterprise edge, thus improving network efficiency. Resilience and throughput are increased, as the enterprise backbone is no longer required to maintain standby routers in back-up mode with un-utilized WAN links. Customers that use Hot Standby Routing Protocol (HSRP) may find they double their site?s available bandwidth without increasing their costs by enabling GLBP.

You can get more information at www.cisco.com/go/grip .

regards,

Rick.

network · ‎05-21-2002

GLBP looks like it solves one problem I've been having with my current land desi

gn layout I've been testing. Using the pop layout above, I've got two vlans act

ing as totally separate ethernet switched networks. Each piece of equipment has

one connection to each vlan, one on each of the two switches in each pop. I'm

doing ospf on both ethernet lans with the gateway router which is also setup on

both vlans. The HP's are doing switchmeshing as discussed, with the mesh carryi

ng both vlans. This allows the system to function should there be a fiber cut,

in which case both vlan's continue pushing data, or switch death, in which case

only the routers with connections on that switch loose one vlan and again, via o

spf, keep pushing data. This schema doesn't work for devices that don't speak o

spf natively however, as they get relegated to one lan or annother with no grace

full fallover means should their switch die.

Would I be able to impliment GLBP and keep the redundancy I've covered above in

place, have all routers/ubrs survive one switch dieing, while eliminating the du

al vlan architeture? (Basically, how will I be able to utilize dual ethernet li

nks for each router, each on a different switch, but on the same 'lan' if I swit

ch to GLBP?) Also, is there any release of IOS that I have access to available

for me to play with this in my lab?

rickwilliams · ‎05-24-2002

I suggest you contact your Cisco account rep and request some information and design assistance. Without really getting into detail on what you need to do and exchanging diagrams etc. it's hard to know if this solution is right for you.

Regards,

Rick.