RE: gigabit router (was Re: Getting a "portable" /19 or /20)
CEF is not the only mechanism to implement distributed forwarding (within or without Cisco for that matter), and to say that distributed forwarding is faulty because of software complexities of one manufacturer, whose code base is built upon on a monolithic core (to use operating system would lower what it means to actually be a operating system) is to generalize all failures to a distributed architecture where the fault does not _always_ lie.
CEF and dCEF periodically break on Cisco. They break in the most interesting ways - the debugging indicates that the packets are going one way when they in reality are going the other way. Without CEF or dCEF Cisco's are useless for the amount of traffic that I am interested in. Juniper is very interesting. The only problem is a gazillion strange things that it tries doing. Juniper's BGP has very intereting bugs in confederations which I have discovered on the day #1 of putting one of them into production. That bug is still not fixed. Since confederations are very widely used, and no one else found this bug, and yet we have hit it nearly immediately, there is some sort of logical problem here. If something as simple as AS_PATH prepending in confederations does not work, I have some big reservations about things that are much more complex than that. Nortel has some exellent hardware that they have inherited from Bay Networks. Unfortunately, they still have not written the software to take any advantage of it, nor they developed it to the level where it can actually complete with the new offerings from the other vendors. Thanks, Alex
While a central architecture is simple, it has been shown within and beyond the industry that it does not scale.
David
-----Original Message----- From: alex@yuriev.com [mailto:alex@yuriev.com] Sent: Wednesday, April 11, 2001 1:41 PM To: nanog@merit.edu Subject: RE: gigabit router (was Re: Getting a "portable" /19 or /20)
Vendors have known how to solve this problem for many years. Failure to do so is a poor implementation and has nothing to do with centralized forwarding being better/worse than distributed forwarding.
Yet another person who does not understand the KISS principle. I am sure in theory it all works great, though I am seeing way too many comments similiar to:
"The connectivity issues have been resolved. This appears to be the same CEF related issues we experienced Monday evening, and we have a case open with Cisco. As we get more information from Cisco, we will be passing it along."
Alex
On Wed, 11 Apr 2001 alex@yuriev.com wrote:
into production. That bug is still not fixed. Since confederations are very widely used, and no one else found this bug, and yet we have hit it nearly immediately, there is some sort of logical problem here. If something as simple as AS_PATH prepending in confederations does not work, I have some big reservations about things that are much more complex than that.
Alex, there are some promising local isps that have large bgp networks with confederations and the confed behavior is fine. I am not sure what you mean by as path prepending in confederations, but if you are talking about the routers not taking into account the confederation member AS's in the overall as path length when doing route selection, than that is what was expected by most people using confeds. [The question of wether AS_PATH prepending of confed member AS's should matter or indeed if as_path length should matter at all is a different question] As for your link state protocol issue, well, that is what we have today that works (of course depends on how you define work, but its worth several billion USD of revenue a year), so I'll take it till something better comes along. /vijay
into production. That bug is still not fixed. Since confederations are very widely used, and no one else found this bug, and yet we have hit it nearly immediately, there is some sort of logical problem here. If something as simple as AS_PATH prepending in confederations does not work, I have some big reservations about things that are much more complex than that.
Alex, there are some promising local isps that have large bgp networks with confederations and the confed behavior is fine. I am not sure what you mean by as path prepending in confederations, but if you are talking about the routers not taking into account the confederation member AS's in the overall as path length when doing route selection, than that is what was expected by most people using confeds.
No, I am talking about Junipers PR#13233. Your friendly Juniper Systems Engineer should be very glad to explain to you what exactly that means. Alex
On Wed, 11 Apr 2001 alex@yuriev.com wrote:
No, I am talking about Junipers PR#13233. Your friendly Juniper Systems Engineer should be very glad to explain to you what exactly that means.
Thanks for sending the PR, and while I don't see a need for it, it may be useful in some cases. Thats not a bug per se though, its an implementation detail ;) /vijay
participants (2)
-
alex＠yuriev.com -
Vijay Gill