They were right - I got three private messages thanking me for pointing out this isn't a Cisco-help list, all accompanied with warnings that someone would react like this. There's a line between hands-on, case-specific config info of use and interest to all providers vs. more basic how-do-I-do-this types of discussions. While Mr. Dale's explanation of CEF load-balancing was accurate and succinct, the same information could've been found by the original poster on CCO or USENET, or even via a quick telephone call to Cisco pre-sales. When I ask case-specific questions here, it's a) because I've done all the research I can do on my own and come up short, and b) because I think the question itself would be one that other operators would care about. I had a lot of questions about MPLS in a network-provider environment a few months back when I converted my core to it (all my previous experience in it was in private networks), but I didn't ask those here because I was able to get the answers elsewhere and I didn't think the other folks on this list would be interested in either the questions or the answers, anyway. The remark wasn't meant in a snide tone, but to point out that there are other resources available for such things, and that it's probably a good idea to try those, first. NANOG should -not- be the first place people ask questions of this type. That answer your question? -----Original Message----- From: John Fraizer To: rdobbins@netmore.net Cc: jaideepc@cs.umn.edu; nanog@merit.edu Sent: 8/8/00 9:38 PM Subject: RE: question about traffic eng/ Cisco CEF. On Tue, 8 Aug 2000 rdobbins@netmore.net wrote:

This question is more appropriately asked in the USENET comp.dcom.* hierarchies.

Why do some people find it necessary to post snide remarks vs posting an informative reply? Don't you think it is possible that someone on NANOG might perhaps know the answer of this question and be able to answer? Was your snide remark intended impose a better S:N ratio in the list? If so, it would have been much better served had you posted it in private and even better had you not posted it at all! Sorry to all the other readers of the list but, I just can't stand this "We're too good to answer your question" bull!

-----Original Message----- From: Jaideep Chandrashekar [mailto:jaideepc@cs.umn.edu]

When I enable cef and set it to share load on a per destination mode, for some reason, all the traffic to a destination takes a single route .. though the route table shows two equal cost paths.

The only point I might be missing is that I am looking at very few connections (2-3) and that load sharing with cef might be engineered to share load in a statistical sense (only kick in with large number of connections).

Could anybody throw any light on this.

I may be wrong but, from what I can remember, what you describe as being the problem is actually it working as designed. Each time a new flow is encountered, it is checked against the current tables. If an interface has not already been assigned to that destination prefix, it is done so then. If you have two total destinations, you're going to end up with one on each of your links. Feel free to correct me if I'm wrong folks. Just please, do it kindly. --- John Fraizer EnterZone, Inc.