Redundant Routes Question - Please Help

[CiscoFun]

PLEASE HELP – Thank you very very much!
At Core Site:
I have two Cisco 7206s each with one DS3 to the frame cloud
Each 7206 is connected to each other through two Cisco 5509 switches via Fast Ethernet

At Remote Site:
Cisco 2610 with two PVCs (one PVC to each of the 7206)
PVC 1 = 64k with 16kcir (DLCI 20 mapped to DLCI 16)
PVC 2 = 64k with 8kcir (DLCI 20 mapped to DLCI 17)

Drawing:

Remote Site (DLCI 20)
----------16kCir---|----8kCir-----------
| |
--------------FRAME CLOUD----------------
| DS3 | DS3
7206 #1 (DLCI 16) 7206#2 (DLCI 17)
| |
5509-------------100Mbit-------------5509


Question:
I am looking at the best way to provide redundancy with these dual DS3.
I am looking into EIGRP however with the CIRs not the same I am not sure this is a good solution.

Any suggests would be a helpful. I am not looking for some to engineer this project for me, I am asking the forum because I am look for advice on what my steps should be. Thank you very much again

 

[l3stuff]

all routing protocols look at links from a cost perpective, not a potential penalty perspective (i.e. CIR). So if the link (tail) bandwith and hop count is the same you'll load balance (important to check you are doing per destination load balancing -CEF default).

The good news, if you use EIGRP and turn down the bw statement, you can still load balance with verience enabled..

 

[CiscoFun]

So if I put "bandwidth 16" and a "bandwidth 8" on the
sub-interfaces at the remote site and on each DS3 sub-interface EIGRP will always take the 16 path unless the it is down, in which case it will take the remaining bandwidth 8 path? (sorry on that long sentance hehe)

On the variance, I would just add "variance 1" in the "router eigrp 1024"?

Sorry if these are silly questions, I have only worked with EIGRP a few time on simple networks.

Thank you for this help!

 

[l3stuff]

The thing to think about is CIR and link bandwidth are not linked here.

EIGRP with CEF will allways proportionately load balance. You have such low CIR's you are (your SP is) probably marking most of traffic DE anyway. Use what you have, just avoid spliting flows across the 2 links

 

[CiscoFun]

Thank you. Will map-class influence routes on the remote and local side?

!
map-class frame-relay 64k8cir
frame-relay cir 64000
frame-relay mincir 8000
frame-relay adaptive-shaping becn
frame-relay bc 1000
!
map-class frame-relay 64k16cir
frame-relay cir 64000
frame-relay mincir 16000
frame-relay adaptive-shaping becn
frame-relay bc 1000
!

Thank you again!

 

[l3stuff]

Routing protocols really look at fairly primitive values on basic link parameters to build reachability info.

So no, sorry 8)

Why don't you want to use whatever bandwidth is there...

 

[CiscoFun]

So i should just put a "bandwidth 16" on the 16 cir etc... ?
or do you mean
Just let it fun at full 64K?

 

[svermill]

Don't confuse traffic shaping with EIGRP path cost. EIGRP will load balance accross two equal cost paths by default. If you set variance to 2 (and include 'traffic-share balanced' in your config) you can load balance proportionally accross your 16k and 8k links. Of course, it is the 'bandwidth xxx' statement on the interface/subinterface that EIGRP looks at when calculating path cost. It is generally considered best to set bandwidth to the true CIR (regardless of what you set the CIR to in your traffic shaping config). The other stuff has to do with traffic shaping and should be coordinated closely with your service provider. For instance, 'frame-relay bc' is determined by the sampling interval of the service provider network. It should probably be some value for your 8k circuit, and twice that value for your 16k circuit. But I have seen enough regional differences in the way that service providers enforce traffic shaping policy to just generally recommend asking them instead of trying to wing it.