Combining two T1 FR ints between two routers

[cseymour-rhs]

I've been looking all over for an answer to this...

I need to combine (load balance but more) two Frame Relay T1 interfaces between two routers.

I would love to use Multilink PPP but seem hampered by the Frame Relay encap. I know CEF and EIGRP will work but order of packets at the destination is important. (Also, I want to NAT all traffic thru these interfaces).

Please, can anyone point me to an example config or doc? I've looked over the few MLPoFR docs but they dont seem to address multiple physical interfaces. Am I missing something?

[ruwhite]

I'm not certain about ppp over frame, but you should be able to load share over two links just using ip routing. If you have more than one set of hosts on either end, then you can use per source/destination address pair, and not worry about out of order packets when load sharing.

Russ.W

[cseymour-rhs]

Yeah, I know what you mean. Problem is I have many users on one side connecting to one app server thru the dual T1s. We are utilizing an ASP thru their private frame relay network. So with only one destination for my users I dont think "per destination" load balancing will work. "Per packet" balancing is process switched (from what I read) so I don't think that's an option. Thanks for the input.

There has to be some way to aggregate two physical frame relay ints. I know about the inverse multiplexers but that's cost prohibitive.

[t.baranski]

Multilink Frame Relay exists, though it's not nearly as mature as Multilink PPP. Search around to see if the hardware/software you're using supports it.

[thisisshanky]

You can run PPP encapsulation over frame-relay using virtual templates. You will specify multilink configurations under the virtual template. While under the frame-relay interface-dlci command you can change the encapsulation to ppp.

int s0.1 point-to-point

frame-relay interface-dlci 100 ppp virtual-template

Check more here....

http://www.cisco.com/warp/public/788/voice-qos/designing-mlp-over-fr-atm.html

[cseymour-rhs]

You are saying...

1. Even though the doc you refer to uses FR at one end and ATM at the other, I can use the same technique with FR at both ends?

2. Even with FR encap at the physical interface, ML PPP on the virtual-template will still be able to aggregate (load balance) the bandwidth over two or more physical ints?

The doc states that there is a requirement of 1 virtual-template int per PVC. Does two physical FR ints mean two PVCs? FYI - I am using the same DLCI on both ints.

Also, note under the 'hints' portion of the doc is says that sub-interfaces are really not necessary.