cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
962
Views
0
Helpful
9
Replies

Load Balancing Options

anderse
Level 1
Level 1

Hello All. I have two sites, each with a 3640 running version 12.1. I have three T1's between the two sites. I want to load balance the three T's. I was looking at Multilink PPP, but our lines are bouncing when using PPP for encapsulation. We have had to switch to HDLC for stability. My question is, what are my options for load-balancing if I have HDLC for encapsulation on all 3 T's?

Thanks all

9 Replies 9

r.sneekes
Level 1
Level 1

You could loadbalancer thru CEF.

You have to enable CEF on both router. And configure cef loadbalancer per packet instead of the default per destination.

You would need three equal cost routes for al routes.

All packets routed thu CEF will then be loadbalanced.

Look at this link for more information

http://www.cisco.com/en/US/products/sw/iosswrel/ps1831/products_configuration_guide_chapter09186a00800ca6ca.html

Eric

The most important part of the answer is to have equal cost routes pointing over each of the three T1s. You can achieve this by using a dynamic routing protocol or by configuring static routes (with a static route for each destination pointing to each of the T1s).

As for whether it is better to use load sharing per packet or per destination I will present a different viewpoint. The main advantage of per packet is that each link would carry equal traffic. The disadvantage of per packet is that it opens the door to out of order delivery of packets. Depending on the application being run, out of order packets may become a significant performance problem. I have seen this happen when per packet load share was specified.

With per destination load share the load on each link may not be exactly the same but each link will be carrying part of the load and it certainly minimizes the possibility of out of order packet delivery. So I would suggest that you use the default of per destination load sharing.

HTH

Rick

HTH

Rick

Thanks Rick. I was wondering if you have an answer as to why my T's won't come up using PPP. My ISP says they have nothing to do with my encapsulation problem. However, when my lines are set to PPP, they bounce all the time and sometimes will not come back up. Setting them to HDLC has been much more stable, but my original goal was to have the 3 T's in a Multilink group.

Eric

I do not yet have enough information to have much opinion about what the problem is for you to use PPP. I do agree that is the links are pretty stable when using HDLC that the instability for PPP is probably not a provider issue. You might want to do a little testing of the links using extended ping. In extended ping specify large frame sizes (1500 or close to it) and specify the payload using the extended options. Do some extended pings with payload with lots of binary zeros (use 0101 as the payload) and with lots of binary ones (use FEFE as the payload).

I am not clear whether all of the links become unstable or if it is one of them. I am not clear whether it becomes unstable when PPP is configured or if it is when multilink is configured. Are there log messages generated when the link becomes unstable and if so what are the messages? It might be helpful if you could run a debug ppp multilink events and/or debug ppp negotiation.

If you can clarify some of these points we may move closer to a solution for your problem.

HTH

Rick

HTH

Rick

Rick

We are using the protocol OSPF for load balancing right now. We want a better solution. We haven't done anything with Multilink yet because we cannot get the T's to come up with PPP encapsulation. When we configure the lines for PPP on each end, we get a status of Line Up and Protocol Down. My ISP says its not there issue, but I do think it has something to do with them. Hope that helps.

Eric

What you describe is very strange. The status of line up line protocol down is caused by failure of keepalives. I am quite puzzled that keepalives would fail for PPP and would not fail for HDLC. It makes me wonder if the connection is not the simple point to point between your two sites that it seems to be.

Would you post the output of show cdp neighbor detail from both routers? Also it might be beneficial to have the router configurations.

HTH

Rick

HTH

Rick

Hi Rick

I wanted to let anyone who reads this thread know that we solved the PPP encapsulation issue. As it turns out, the problem had to do with our ACS server. AAA statments were put into the routers that caused the routers to try and authenticate with our ACS server for the PPP connections. We did not have any authorization setup on the ACS server as it was in place purely for our wireless AP's. I removed the statements in the config for PPP authentication to the ACS server as well as network authorization through the ACS server and the lines all came up.

Eric

Thanks for posting the update and letting us know about the resolution of this issue. It is very helpful to know when things have been resolved and what the resolution was. It is especially helpful in situations like this which are quite out of the ordinary. Perhaps this will be helpful to other people who get proplems using PPP. Perhaps if the config had been posted we might have noticed the authentication - but who knows.

Congratulations on getting your issue resolved.

HTH

Rick

HTH

Rick

truhje
Level 1
Level 1

ip load-sharing

To enable load balancing for Cisco Express Forwarding (CEF), use the ip load-sharing command in interface configuration mode. To disable load balancing for CEF, use the no form of this command.

ip load-sharing [per-packet] [per-destination]

Ref: http://www.cisco.com/univercd/cc/td/doc/product/software/ios121/121cgcr/switch_r/xrdscmd2.htm#wp1063377

:o) Truls