Comments inline ""Michael Williams"" wrote in message [EMAIL PROTECTED]">news:[EMAIL PROTECTED]... > nrf wrote: > > If you are using PPP authentication, then it will use the > > authenticated name > > of the other side to determine what link goes with what > > bundle. It can also > > use a unique identifying value called the > > endpoint-discriminator, which is a > > a number that is negotiated when multilink is first negotiated > > in LCP. It > > can use both authenticated name and endpoint discriminator, and > > this is the > > default behavior. This behavior can be controlled using > > "multilink > > bundle-name" commands. > > We're not using PPP authentication on the links. So it must be using the > endpoint discriminator. Do you know of a webpage or book that lays out > specifically the process that Cisco routers use to determine which links go > in which bundles?
I don't know where to find the specific Cisco implementation of the endpoint discriminator. But the concept is described in RFC 1990 in detail. I think the RFC should have what you need. Whenever in doubt, always consult the RFC's. And you can watch for IOS to negotiate the discriminator when you debug ppp negotiation. > > > In my experience, doing it the with direct MLP bundle command > > is flakier. > > I've had several weird multilink problems that I solved just > > reverting to > > the virtual-template method. The virtual-template way has > > given me less > > problems. But the direct MLP way is more intuitive. > > > > But as a side-note, I often see MPPP on serial interfaces as > > sort of a > > fool's errand, unless perhaps you are linking a Cisco router to > > another > > vendor (MPPP on dial networks like ISDN is still useful). The > > only really > > good reason to use MPPP on a serial link is for link > > load-balancing. But if > > you just need link load-balancing, then all I have to say is > > "CEF". End of > > story. CEF accomplishes the same thing, and is overall > > faster, less > > resource-intensive, and more stable than MPPP. > > Well, we considered using EIGRP load balancing, but I can't find > verification that EIGRP will load balance IPX traffic, which is needed. We > run both IP and IPX. As far as CEF goes, the routers at the remote ends of > our bundles are uslaly 2500s, so I don't think CEF is an option. We are > using MLP to load balance over multiple T1s. The virtual-templates seems to > be working just fine, so I guess my questions here are more academic in > nature. Yes, EIGRP will load-balance IPX traffic. Of course, it suffers from the same problem with load-balancing IP fast-switched packets, unless you're willing to just use IPX process switching (a painful solution). > > Thanks for your input! > Mike W. Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=27164&t=26877 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
