Just to clarify - I never said RT was part of the bestpath selection. I did
put the "iPhone disclaimer" (twice, actually because I knew someone is
going to jump at the short explanation), which means I was on the road and
the example may not have been the clearest possible. Now I'm not on iPhone,
so allow me to clarify my original answer, by building on yours and
Narbik's.
What you wrote is not by a long shot incorrect, but it's also not the exact
reason *why* RTs are not used as a differentiator or *why* we have RDs and
RTs to begin with, which is something many students struggle with. We all
know what they do, but I like to explain *why* certain things behave the
way do.
There are two reasons for having address space expanded from the original
32 bits:
1. Allow for overlapping addresses in different VPNs
2. Allow for iBGP load-sharing within the same VPN
The main reason why the route distinguisher is *not* used to associate
routes with VPNs is the lack of flexibility that would provide. Basically,
if we used only RD for VPN membership in a fashion similar to
route-targets, the possible logical topologies would be limited without
extensive filtering (anything is doable with the ample amount of
route-maps). Using communities to associate routes with VPNs allows for a
more architectural and scalable approach to route association with VPNs
than using a single entity like RD ever does.
Thus, we have two different entities we have to deal with: RD and RT and
a fundamental reason for having a separation between RD and RT is to allow
for topology flexibility and uniqueness of prefixes. The traditional book
example you gave with overlapping customer addresses is just one of the
reasons for RDs. Another, also very common is the inability to use iBGP
load-balancing when using same RDs within the same VPN, when using
route-reflectors. The reason is, of course, that RRs advertise only the
best prefix. I believe there are some recent RFCs that address this issue,
but I'm yet to see the working code.
Narbik: VRFs work just fine without RDs. It's MPLS VPNs that don't and
there is a difference between the two :-)
-- Marko Milivojevic - CCIE #18427 (SP R&S) Senior CCIE Instructor - IPexpert -- Marko Milivojevic - CCIE #18427 (SP R&S) Senior CCIE Instructor - IPexpert On Mon, Mar 26, 2012 at 18:27, Brian McGahan <bmcgahan_at_ine.com> wrote: > Personally that seems overly confusing to me. Yes Route Targets are an > attribute of the route, but that attribute is not part of the BGP Bestpath > Selection. I'm not sure how it ties together. It's simpler to think of it > this way: > > It's given that customers of a Service Provider will have overlapping IP > addressing in their VPNs, e.g. you will have more than two customers who > use the 10.0.0.0/8 network. The RD is how you tell them apart. If you > have customer "A" with RD "A" and customer "B" with RD "B" the routes "A: > 10.0.0.0/8" and "B:10.0.0.0/8" become unique. This is all the RD does. > > The Route Target tells you which VRF table the route belongs to. You have > to separate the two attributes because sometimes you want the same route to > belong to multiple VRF tables. This is common in what's known as "Central > Services VPNs". For example if the Service Provider hosts email for > customers, that route to the mail server would have to be in the routing > table of multiple customers. This doesn't break the rule of the route > having to be unique though, which is what the RD does. > > Like I said you may be able to find more clarification in this video: > http://goo.gl/Y0imB. > > Brian McGahan, CCIE #8593 (R&S/SP/Security) > bmcgahan_at_INE.com > > Internetwork Expert, Inc. > http://www.INE.com > > -----Original Message----- > From: nobody_at_groupstudy.com [mailto:nobody_at_groupstudy.com] On Behalf Of > Marko Milivojevic > Sent: Monday, March 26, 2012 6:00 PM > To: Yemi Salau > Cc: ccielab_at_groupstudy.com > Subject: Re: OT: GS Archives Search > > Simple reason - prefixes are passed on through the bestpath selection > process where the best one is chosen based on attributes. RT is a > community, which is an attribute. This means that given two prefixes with > different RTs would be treated as equals when it comes to bestpath > selection. With RD we extend the prefix space to 86 bits and then use those > for comparison instead of 32bit ones. > > [ iPhone, brevity, etc disclaimer :-) ] > > -- > Marko Milivojevic - CCIE #18427 > > :: This message was sent from a mobile device. I apologize for errors and > brevity. :: > > On Mar 26, 2012, at 14:44, Yemi Salau <salauolayemi_at_yahoo.co.uk> wrote: > > > Thanks Marko, reading RFC 4364, I was trying to figure out why RT > > couldn't > do the same job of RD for uniquely separating VPN-IPv4 routes within the > provider MPLS cloud. I'll watch your video when I get home. Many Thanks. > > > > From: Marko Milivojevic <markom_at_ipexpert.com> > > To: Yemi Salau <salauolayemi_at_yahoo.co.uk> > > Cc: ccielab_at_groupstudy.com > > Sent: Monday, 26 March 2012, 16:25 > > Subject: Re: OT: GS Archives Search > > > > > > Yemi, > > > > I'm not sure about the Archive search, but I can certainly help you > > with RD > and RT. Almost two years ago I hosted a free online training session > called "MPLS 101". You can find it, together with all other recorded > vLectures on many other subjects here: > > > > http://bit.ly/vLecture > > > > Please go ahead and watch it and if you have any questions feel free > > to ask > them here. > > > > -- > > Marko Milivojevic - CCIE #18427 (SP R&S) Senior CCIE Instructor - > > IPexpert > > > > On Mon, Mar 26, 2012 at 10:18, Yemi Salau <salauolayemi_at_yahoo.co.uk> > wrote: > > Guys, > > > > I remember a time where I was able to search the GS archives for > > stuffs. Is > this still available today? I want to search out some stuffs on RD vs RT. > > > > Yemi > > > > > > Blogs and organic groups at http://www.ccie.net > > > > ______________________________________________________________________ > > _ Subscription information may be found at: > > http://www.groupstudy.com/list/CCIELab.html > > > Blogs and organic groups at http://www.ccie.net > > _______________________________________________________________________ > Subscription information may be found at: > http://www.groupstudy.com/list/CCIELab.html Blogs and organic groups at http://www.ccie.netReceived on Mon Mar 26 2012 - 19:42:45 ART
This archive was generated by hypermail 2.2.0 : Sun Apr 01 2012 - 07:56:52 ART