The MPLS-OPS Archive

Cell Relay Retreat>MPLS-OPS Archive>month:2003-Feb> msg00106



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

Re: Route Distinguisher Questions

  • From: "wanglei" <wanglei@harbournetworks.com>
  • Date: Sun, 16 Feb 2003 21:21:46 +0800
  • Resent-Date: Sun, 16 Feb 2003 10:03:55 -0500
  • To: <MHeusinger@gmx.net>, "Aries C" <cariesc@hotmail.com>, <mpls-ops@mplsrc.com>
  • X-MIME-Autoconverted: from base64 to 8bit by host.secure4-hosting.net id h1GDNLL04348

Martin's explaination is so good ^_^

----- Original Message ----- 
From: "Martin Heusinger" <MHeusinger@gmx.net>
To: "Aries C" <cariesc@hotmail.com>; <mpls-ops@mplsrc.com>
Sent: Sunday, February 16, 2003 8:29 PM
Subject: Re: [MPLS-OPS]: Route Distinguisher Questions


> Hi Aries,
> 
> to answer your questions:
> 
> Route Distinguisher (RD) has only one purpose: make IPv4 prefixes globally 
> unique. MPLS VPNs are based on the distribution of routing information of 
> IPv4 prefixes. Now customers may all use private IP addresses (RFC 1918). 
> To be able on a PE to distinguish between f.e. 10.0.0.0/8 from one customer 
> to 10.0.0.0/8 of another customer one adds the Route Distinguisher. In 
> Multiprotocol BGP the PEs are routing VPNv4 prefixes, i.e. VPNv4=RD+IPv4. 
> There is no other special meaning to the RD than to make addresses globally 
> unique. Especially RD has nothing to do with VPN membership.
> 
> VPN membership is solely determined through import and export rules based 
> on Route Targets (RT). Those are extended communities attached to VPNv4 
> prefixes. One VPNv4 prefix may have one or many different RTs attached to 
> it, i.e. a IPv4 prefix may be seen in one VPN or in many VPNs or customer 
> locations.
> 
> In a first approach you might say an RT identifies a VPN. A customer site 
> is part of several VPNs, when it´s IPv4 prefixes (f.e. 10.0.0.0/8) can be 
> reached within several VPNs. This CAN be achieved by attaching the 
> respective VPN-RTs to the VPNv4 prefixes in every PE a customer site is 
> connected to.
> (I am aware, that there might be different usages for RT, like in central 
> services VPN, but as an MPLS instructor my experience is, that this 
> approach is the most simple and understandable one "for the beginner").
> 
> You can think of VRFs as a virtual Router the customer Edge router (CE) is 
> attached to. Within one PE very VRF has to have a unique RD. IF a customer 
> has two CEs (CE1, CE2)with the same connectivity requirements connected to 
> the same PE you could use ONE VRF in the beginning. This VRF would have a 
> routing table with all the IPv4 prefixes to be seen by the two customer 
> sites.
> Now assume a more complex VPN topology. F.e. CE1 might be part of 
> additional VPNs, whereas CE2 is not. This clearly means, that CE1 should 
> see a different set of IPv4 prefixes than CE2. This means one needs TWO 
> VRFs one for CE1 and one for CE2. So there have to be two RDs as well. 
> Starting with the simple approach of one RD for all VRFs a customer is 
> attached to, this means to create a new VRF (with a unique RD) and to 
> reconfigure the PE. This leads to a network outage. for at least one CE.
> 
> 
> now on to your questions:
> 
> 1. is definition RD per VRF : all RDs are unique regardless which VPN it 
> belongs to ?
> Yes, this is the most flexible setup. VPN membership can be easily changed 
> by adding or removing RTs
> 
> 2. Is definition RD per VPN : one RD is used acrossed all VRF belongs to a 
> VPN ?
> Yes, this can be used in a "simple VPN" scenario, i.e. all customer sites 
> have connectivity to each other.
> 
> 3. Why the paragraph mentions RD per VRF is more recommended ( because 
> configuration conflict and reduce the memory requirements), as compared to 
> RD per VPN ?
> The answer is: RTs are BGP communities. Changing/manipulating them is 
> rather simple. To change the RD (in Cisco routers) one has to delete the 
> VRF and to add it again. This means a lot of work and network outage.
> 
> 4. Can I use RD per VPN, though a site could belong to multiple VPN. 
> Because theoritically, we can use Route Target to control export and import 
> (again the uniqueness is taken care with RD, and also different VPN has 
> different RD anyway i.e. VPN A = RD 100:27 and VPN B = 101:27).
> 
> No. VPN and RD have nothing to do with each other. Also PLEASE: the 
> official RD format is <AS number>:<any number (32 Bit)> or <official 
> IP>:<any number (16 Bit)>!! Using any different numbering scheme is like 
> using "unofficial" IP addresses! DO NOT DO THIS!
> Use 100:27 and 100:28 IF AND ONLY IF your official AS number is 100.
> 
> The confusion about RD and RT and their usage is unfortunately pushed by 
> the fact, that RT has the same format.
> 
> Hope this helps more than it confuses ;-)
> 
> Martin
> 
> P.S.: The book is right, Ivan understands the matter :-)
> 
> On Sun, 16 Feb 2003 08:58:10 +0000, Aries C <cariesc@hotmail.com> wrote:
> 
> > Hi All,
> >
> > I m reading a book of Ivan Pepelnjak's book (MPLS and VPN Arch.)
> > At page 191 Chapter 10, there's 3 paragraph get me confused :
> >
> >
> > "Each VRF within the PE router configuration needs to have an associated 
> > RD, which might or might not be related to a particular site or VPN 
> > membership of that site. In the most common case, where a site belongs 
> > only to one intranet VPN, it's technically possible and recommended, to 
> > use unique RD for the VPN. However, if this site at some point in the 
> > future will become a member of an extraner VPN, do not take this approach 
> > because it might incur config. issues when trying to provision the 
> > extranet VPN.
> >
> > For example, suppose a different RD is used for each VPN. If a particular 
> > site wants to be a member of multiple VPNs, it's not possible to to 
> > identify which route distinguisher to use for the site because it belongs 
> > to more than one VPN.
> >
> > Therefore, for network topologies other than simple intranet model, use 
> > the same route distinguisher per VRF, rather than per VPN, to avoid this 
> > type of conflicting config. and to reduce the memory requirements of the 
> > PE router. In the case of an extranet VPN, this means the VRF that makes 
> > up the VPN uses the same RD regardless of the particular VPN site to 
> > which the VRF belongs. "
> >
> >
> > The question are :
> >
> > 1. is definition RD per VRF : all RDs are unique regardless which VPN it 
> > belongs to ?
> > 2. Is definition RD per VPN : one RD is used acrossed all VRF belongs to 
> > a VPN ?
> > 3. Why the paragraph mentions RD per VRF is more recommended ( because 
> > configuration conflict and reduce the memory requirements), as compared 
> > to RD per VPN ?
> > 4. Can I use RD per VPN, though a site could belong to multiple VPN. 
> > Because theoritically, we can use Route Target to control export and 
> > import (again the uniqueness is taken care with RD, and also different 
> > VPN has different RD anyway i.e. VPN A = RD 100:27 and VPN B = 101:27).
> >
> >
> > Could anyone educate me on this.
> >
> >
> > TIA
> >
> > Aries C.
> >
> >
> >
> >
> >
> > _________________________________________________________________
> > Hotmail now available on Australian mobile phones. Go to  
> > http://ninemsn.com.au/mobilecentral/hotmail_mobile.asp
> >
> > -------
> > The MPLS-OPS Mailing List
> > Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
> > Archive: http://www.mplsrc.com/mpls-ops_archive.shtml
> >
> >
> 
> 
> 
> 
> -------
> The MPLS-OPS Mailing List
> Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
> Archive: http://www.mplsrc.com/mpls-ops_archive.shtml
> 

-------
The MPLS-OPS Mailing List
Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
Archive: http://www.mplsrc.com/mpls-ops_archive.shtml