[cisco-bba] load balancing over parallel DSL lines (fwd)

[Mark E. Mallett]

I sent this to the cisco-nsp list last week.  Figured I'd
resend it to cisco-bba (even though I imagine there is a
wide overlap).

Yours,
-mm-


----- Forwarded message from "Mark E. Mallett" <mem at mv.mv.com> -----

> Subject: [nsp] load balancing over parallel DSL lines
> From: "Mark E. Mallett" <mem at mv.mv.com>
> Date: Thu, 6 Nov 2003 14:44:37 -0500
> To: cisco-nsp at puck.nether.net
> 
> Hi-
> 
> Wondering if anyone has success in doing load balancing over DSL
> destinations backhauled via ATM.  Given RBE setup including this:
> 
> ip cef
> !
> interface atm1/0
> !
> interface atm1/0.301 point-to-point
>  ip unnumbered loopback 1
>  atm route-bridge ip
>  pvc dsl-301 1/301
>   encapsulation aal5snap
> !
> interface atm1/0.302 point-to-point
>  ip unnumbered loopback 1
>  atm route-bridge ip
>  ip load-sharing per-packet
>  pvc dsl-302 1/302
>   encapsulation aal5snap
> 
> and two DSL circuits to the same customer with two different IP
> addresses ip1 and ip2 on different subnets, I'd like to route
> their subnet to those two destinations and have per-packet load
> balancing occur.  e.g:
> 
>    ip route <subnet> <mask> <ip1>
>    ip route <subnet> <mask> <ip2>
> 
> (I'm using shorthand instead of real numbers for attempted clarity)
> 
> Both ip1 and ip2 are routed over the same pvc (in this case, the 302
> PVC).  
> 
> What I see is that the subnet routing goes via only one of the circuits.
> And indeed:
> 
> gw> show ip cef <subnet> internal
> <subnet>/xx, version 94, per-packet sharing
> 0 packets, 0 bytes
>   via <ip1>, 0 dependencies, recursive
>     traffic share 1, current path
>     next hop <ip1>, ATM1/0.302 via <ip1>/32
>     valid glean adjacency
>   via <ip2>, 0 dependencies, recursive
>     traffic share 1
>     next hop <ip2>, ATM1/0.302 via <ip2>/32
>     valid glean adjacency
>   0 packets, 0 bytes switched through the prefix
>   tmstats: external 0 packets, 0 bytes
>            internal 0 packets, 0 bytes
>   Load distribution: 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 (refcount 1)
> 
> I have tried a number of different options, including enabling/disabling
> various sorts of route cache and applying "ip load-sharing per-packet"
> to all pvcs. 
> 
> Any insight would be appreciated.
> 
> Yours,
> -mm-