Load sharing between full T1 and fractional T1

I've got a new part of T1s from our office to an office in a different city. We have one T1 that's completely for data, and another one that's

14 channels for data, and the rest is used for voice.

How do I share the load in the most efficient way? Right now, I have two equal cost static routes on each side. The fractional T1 seems to be getting a fair bit more of the traffic, which I didn't expect. Is there any way without additional hardware to get the full T1 to handle a bit more of the traffic than the other line?

I've tried putting in the 'bandwidth xxx' statements on each serial interface, but I didn't know if that was just informational, or if that affected the operation in any way.

Thanks!

Reply to
srp336
Loading thread data ...

Trying using EIGRP unequal cost laod balancing

see

formatting link

Reply to
Merv

The bandwidth statements do not affect (significantly) routing, their primary impact is on calculation of the interface loading when you do a show interface.

Since you are looking for approximately 2 to 1 for your load balancing, you could use EIGRP unequal cost load balancing, or you could add another static route for your full T1 link so that the routers think there are three routes between the two sites rather than just two, and the resulting three way load balance gets you closer to the 2 to 1 you are looking for.

You could also approach the challenge using policy routing or longer prefix routing if you have some applications which are more important than others and want to make sure they get adequate bandwidth. Unfortunately, there are a wide variety of ways to split up the traffic and which is best will strongly depend upon your specific requirements.

Good luck and have fun!

Reply to
Vincent C Jones

Thanks for the replies... I'll check into those things. I had a question... do I want per-destination (which I assume is the default) or per-packet load sharing? There probably won't be any VoIP or video streaming or anything like that...

I don't think there's any applications that are more important than others (for right now at least). I think the biggest concern is to allow all available bandwidth to be used when needed.

Thanks again!

Reply to
srp336

You have a choice:

Select "Per destination" and suffer from misbalancing whenever the high persistence flows happen to favor one or the other link.

Select "Per Packet" and suffer from poor applications performance as the packets get out of order because a packet sent via the fast link can arrive before the previous packet sent via the slow link.

Select "MLPPP" and watch your CPU utilization go through the roof. (Test carefully, while MLPPP theoretically can handle this, my memory is foggy about how well Cisco implements MLPPP across mismatched links--probably depends upon the IOS release).

Good luck and have fun!

Reply to
Vincent C Jones

~ ~ Select "MLPPP" and watch your CPU utilization go through the ~ roof. (Test carefully, while MLPPP theoretically can handle this, ~ my memory is foggy about how well Cisco implements MLPPP across ~ mismatched links--probably depends upon the IOS release). ~ ~ Good luck and have fun!

If you have 28/37/3800 routers (and even some lesser models), MLPPPing a bundle containing two T1 or below links should be no problem. Try it, you might like it.

IOS MLPPP can do good load sharing across unequal cost links. Just be sure to set the (surprise!) "bandwidth" value correctly on each member physical interface. I wouldn't include links that vary by more than a 1:3 ratio, but with a 14:24 ratio, you should be good.

Cheers,

Aaron

Reply to
Aaron Leonard

There's a 4500 on our end and a 2621 on the far end. Would it be worth it to try mlppp? How bad would the cpu utilization get?

Reply to
srp336

Cabling-Design.com Forums website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.