The situation I worked with was a service provider (Latin America or East Asia; I worked both and I don't remember which). They had 5 cities, with FR connections between them, but not a full mesh. What they wanted (and I did not fully understand their rationale, but that wasn't a requirement) was for City A-->City C via City B, but City C-->City A via City D or E. Even though there was obviously a bidirectional physical link from each city to the city at the other end of the wire, on a logical basis, A-->C was over a different physical routing than the return traffic form C-->A.
In planning for provisioning the switches, we had "fun" establishing a DLCI numbering convention to reflect physical endpoints and other info re the link (like a priority or QoS indicator) so that seeing the DLCI number would tell you things you needed to know. For instance, the DLCI used from A-->C might have priority 2, so its number would be 132 (1=A, 3=C, + priority). We had to worry about reserved DLCI numbers, etc. We also had to set up the "directionality" of the forwarding to be sure that the intended data path was the one which would be forced by mapping incoming DLCI 132 to outgoing DLCI 342 on the intermediate switches, e.g. I think their rationale for the seemingly funny routing had to do with asymmetric loading on the traffic directionality (client vs. server sort of thing, client makes a simple request, server replies with lots of data). >From a logical perspective, the data conversation was unidirectional due to the separate paths taken from client to server and back. FR provisioning seems in lab environments to be always symmetrical: AB, but it doesn't have to be in the real world. The latter can be achieved by steering data to a given DLCI, which goes to a location other than the obvious one. As I recall, it was an ugly project. Annlee ""Zsombor Papp"" wrote in message news:[EMAIL PROTECTED] > At 10:21 PM 6/24/2003 +0000, MADMAN wrote: > >The spokes only connect via the hub if you don't have a PVC between > >the spokes. It doesn't matter if your uni or bi, > > You can actually buy unidirectional PVC service? What do people do with that? > > Curious, > > Zsombor > > > you have 2 DLCI's per > >PVC, one on each end. In our frame network we using local addressing of > >DLCI's, DLCI 16 could be on both ends of a PVC. Some carries use what > >is called global which I think you may be refering to. The users don't > >have control of the DLCI numbering if connecting to the public frame > >network but can request DLCI's which we can usually accomodate. I have > >no idea what a "forward" and "return" DLCI is!?! > > > > Dave > > > >annlee wrote: > > > Even if it is "switched from spoke-to-spoke", at Layer 1 the spokes > connect > > > via the hub. And to do anything with the traffic, Layer 2 must be > > > consulted -- which gives us Priscilla's DLCI switching table. And, unless > > > the traffic is unidirectional, you will need DLCIs for the opposite > > > direction, as well. I don't know Cisco FR that well, but in at least some > > > vendors' FR implementations, the "return" DLCIs do not have to have the > >same > > > numbers as the "forward" ones. That actually enables you to number > >according > > > to a pattern which indicates connectionality. And it also makes the DLCI > > > switching table twice the size that Priscilla showed. > > > > > > Annlee > > > > > > ""Larry Letterman"" wrote in message > > > news:[EMAIL PROTECTED] > > > > > >>My opinion is that it will go to the hub site since it's a point to > > >>point network.. > > >>If the hub were to be a multi-point connection to the spokes, which > > >>would be one network, > > >>Then the traffic could be switched from spoke to spoke... > > >> > > >> > > >>Larry Letterman > > >>Cisco Systems > > >> > > >> > > >> > > >> > > >>-----Original Message----- > > >>From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of > > >>Aaron Ajello > > >>Sent: Tuesday, June 24, 2003 10:06 AM > > >>To: [EMAIL PROTECTED] > > >>Subject: FR concept question [7:71263] > > >> > > >> > > >>This is probably a very simple concept question, but I've asked a couple > > >>people and haven't gotten a solid answer. > > >> > > >>If I've got two frame relay spoke sites connected point to point with a > > >>hub site and a server in one spoke site copies a file to a server in the > > >>other spoke site, does all the traffic pass through the hub site, or is > > >>it switched within the frame cloud? > > >> > > >>I guess what I'm wondering is does a frame cloud act somewhat like a > > >>lan, where initially packets will go through the default gateway and be > > >>routed and then the following packets will be switched? > > >> > > >>thanks. > >-- > >David Madland > >CCIE# 2016 > >Sr. Network Engineer > >Qwest Communications > >612-664-3367 > > > >"Government can do something for the people only in proportion as it > >can do something to the people." -- Thomas Jefferson Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=71310&t=71263 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]