On Wed, 6 Oct 1999, Jim Warner wrote:
> The University's Astronomical Observatory is connected back to the
> main campus by T-1 line(s) from Pacific Bell. Til now, we've shared a
> single T-1 with our campus phone system on a 1 Mbit for data, 500K for
> the phones basis.
>
> The data guys want (surprise) more BW. One option would be to give a
> full T-1 to the data side by getting a second T-1. That would leave
> about 1 Mbit/sec unused on the "telco" T-1. Or we could leave the
> current setup more or less intact and add a second T-1 for data.
> T-1s on this path cost us about $400 per month.
>
> Typical of scientific computing, the users probably want to be able to
> use their full datacomm capabilities on a single TCP connection.
> Schemes that select a path based on a hash of IP addresses would
> probably be ineffective.
>
> The end point routers are 2500s, but they could turn into 2600s if
> that was important. I'm inclined to think that MPP (multilink PPP)
> might be a good solution. What I don't know is how it would deal with
> a moderate imbalance in the speeds of the links it was asked to merge
> -- 1 Mb/s +1.5 Mb/s. Does anyone have any experience with that? Any
> other suggestions? VoIP and sending the PBX to a dark place is not
> currently an option.
>
>
> -jim warner, UC Santa Cruz
Try this:
! Router A
interface Serial 0
description Partial T1 to Router B
ip address 192.168.1.1 255.255.255.252
ip load-sharing per-packet
no ip route-cache cef
no ip route-cache optimum
no ip route-cache distributed
load-interval 30
no fair-queue
down-when-looped
transmit-buffers backing-store
hold-queue 250 in
!
interface Serial 1
description Full T1 to Router B
ip address 192.168.1.5 255.255.255.252
ip load-sharing per-packet
no ip route-cache cef
no ip route-cache optimum
no ip route-cache distributed
load-interval 30
no fair-queue
down-when-looped
transmit-buffers backing-store
hold-queue 250 in
!
ip route 10.0.0.0 255.0.0.0 Serial 0
ip route 10.0.0.0 255.0.0.0 Serial 1
Because you are using a 2500 on each end, you may not want to turn on "ip
load-sharing per-packet" or turn off "ip route-cache". This would bring
the load up to about 95% on your routers.
The main things to look at here are the "down-when-looped" statements and
the routes to the Serial interfaces instead of the remote-side IP
addresses. I believe the type of effect achieved here would be that one
T1 would fill up and then the other.
The rest of the stuff, like hold-queue, etc isn't really neccessary
either.
Evan
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:06 EDT