MFR or MLPPP?

[sschingel]

I have 2 point to point T1's going from my data center to a larger branch office. At the data center, they are on a multi-channel DS3 on a 7204vxr. At the branch office, I have a 2620 with 2 T1 WIC's. We recently upgraded our internet connection to a 3xT1, and they used MFR to group the 3 T1's together. However, my T1's to the branch office are NOT frame relay, so I didn't know if MFR would work. I saw a post that mentioned MLPPP and thought that might be useful, but couldn't find any config examples. Opinions?

[tdrais]

Unless you have a issue with load balancing I would not run either. With CISCO on both ends load balance by packet using CEF normally does a ok job. Your main reason to use one of the others is if you have a large data stream that attempts to use more than 1 circuits bandwidth.

The other reason to not use either of these is they put more load on both the router to encapsulate the data. You will also get slightly less capacity since there is overhead associated with both MFR and MLPPP.

In your case you will most likely have to use MLPPP if you need . MFR in most cases requires a frame switch that supports this option.

[sschingel]

Thanks. I do have a slight issue with load balancing. The way it's set up now, a conversation only uses one of the T1's. At night, we are doing a backup, and it is only using 1 of the T1's instead of both. So it is not finishing. I have the statement "ip cef" in the config so I know it's using both T1's during the day on an alternating basis, but I was hoping to get a full 3Mb/sec throughput like we have with our internet 3xT1.

[tdrais]

You want to put

ip load-sharing per-packet

in each t1 interface

The default cef load balance is per destination so you get poor utilization.

This document talks about the 2 ways CEF sends traffic.

http://www.cisco.com/en/US/products/hw/modules/ps2033/prod_technical_reference09186a00800afeb7.html

[sschingel]

That seems to have worked - in one direction. Traffic from the data center to the remote site seems to be using both of the T1's. However, traffic from the remote site to the data center still seems to be alternating. I am attaching 3 files. One has the configs for the serial interfaces for both routers. The other 2 are MRTG printouts showing the data patterns. Let me know if there's anything else I should change. Thanks...

[tdrais · Accepted solution]

Kinda hard to say. Your configuration is correct and it should be working.

The only guess I would have is that you are seeing the differences in packet size. When you have a very low volume of traffic it can make the rates look funny. It could be getting unlucky and sending more large packets on one circuit than the other.

I would try to transfer some larger files from the remote site to the local site and see if your graph looks better. I suspect everything is ok.