Internet Draft
Network Working Group                                      Eric C. Rosen
Internet Draft                                             Yakov Rekhter
Expiration Date: January 2001                        Cisco Systems, Inc.

Tony Bogovic                                        Stephen John Brannon
Ravichander Vaidyanathan                           Monique Jeanne Morrow
Telcordia Technologies                                       Swisscom AG

Marco Carugi                                        Christopher J. Chase
France Telecom                                                       ATT

Ting Wo Chung                                           Jeremy De Clercq
Bell Nexxia                                                      Alcatel

Eric Dean                                                   Paul Hitchin
Global One                                                  Adrian Smith
                                                                      BT

Manoj Leelanivas                                           Dave Marshall
Juniper Networks, Inc.                                          Worldcom

Luca Martini                                            Vijay Srinivasan
Level 3 Communications, LLC                        CoSine Communications

Alain Vedrenne
SITA EQUANT

                                                               July 2000


                             BGP/MPLS VPNs


                     draft-rosen-rfc2547bis-02.txt

Status of this Memo

   This document is an Internet-Draft and is in full conformance with
   all provisions of Section 10 of RFC2026.

   Internet-Drafts are working documents of the Internet Engineering
   Task Force (IETF), its areas, and its working groups.  Note that
   other groups may also distribute working documents as Internet-
   Drafts.

   Internet-Drafts are draft documents valid for a maximum of six months
   and may be updated, replaced, or obsoleted by other documents at any
   time.  It is inappropriate to use Internet-Drafts as reference



Rosen, et al.                                                   [Page 1]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


   material or to cite them other than as "work in progress."

   The list of current Internet-Drafts can be accessed at
   http://www.ietf.org/ietf/1id-abstracts.txt.

   The list of Internet-Draft Shadow Directories can be accessed at
   http://www.ietf.org/shadow.html.


Copyright Notice

      Copyright (C) The Internet Society (2000).  All Rights Reserved.


Abstract

   This document describes a method by which a Service Provider may use
   an IP backbone to provide VPNs for its customers.  MPLS is used for
   forwarding packets over the backbone, and BGP is used for
   distributing routes over the backbone.  The primary goal of this
   method is to support the case in which a client obtains IP backbone
   services from a Service Provider or Service Providers with which it
   maintains contractual relationships.  The client may be an
   enterprise, a group of enterprises which need an extranet, an
   Internet Service Provider, an application service provider, another
   VPN Service Provider which uses this same method to offer VPNs to
   clients of its own, etc.  The method makes it very simple for the
   client to use the backbone services.  It is also very scalable and
   flexible for the Service Provider, and allows the Service Provider to
   add value.

   This document obsoletes RFC 2547.



















Rosen, et al.                                                   [Page 2]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000




Table of Contents

    1          Introduction  .......................................   4
    1.1        Virtual Private Networks  ...........................   4
    1.2        Edge Devices  .......................................   5
    1.3        Multiple Forwarding Tables in PEs  ..................   6
    1.4        VPNs with Overlapping Address Spaces  ...............   6
    1.5        VPNs with Different Routes to the Same System  ......   7
    1.6        SP Backbone Routers  ................................   7
    1.7        Security  ...........................................   8
    2          Sites and CEs  ......................................   8
    3          VRFs: Per-Site Forwarding Tables in the PEs  ........   9
    4          VPN Route Distribution via BGP  .....................  11
    4.1        The VPN-IPv4 Address Family  ........................  11
    4.2        Encoding of Route Distinguishers  ...................  12
    4.3        Controlling Route Distribution  .....................  13
    4.3.1      The Route Target Attribute  .........................  13
    4.3.2      Route Distribution Among PEs by BGP  ................  15
    4.3.3      Use of Route Reflectors  ............................  16
    4.3.4      How VPN-IPv4 NLRI is Carried in BGP  ................  19
    4.3.5      Building VPNs using Route Targets  ..................  19
    5          Forwarding Across the Backbone  .....................  20
    6          Maintaining Proper Isolation of VPNs  ...............  21
    7          How PEs Learn Routes from CEs  ......................  22
    8          How CEs learn Routes from PEs  ......................  25
    9          Carriers' Carriers  .................................  25
   10          Inter-Provider Backbones  ...........................  26
   11          Accessing the Internet from a VPN  ..................  28
   12          Management VPNs  ....................................  30
   13          Security  ...........................................  31
   14          Quality of Service  .................................  31
   15          Scalability  ........................................  32
   16          Intellectual Property Considerations  ...............  32
   17          Acknowledgments  ....................................  33
   18          Authors' Addresses  .................................  33
   19          References  .........................................  36
   20          Full Copyright Statement  ...........................  37












Rosen, et al.                                                   [Page 3]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


1. Introduction

1.1. Virtual Private Networks

   Consider a set of "sites" which are attached to a common network
   which we may call the "backbone". Let's apply some policy to create a
   number of subsets of that set, and let's impose the following rule:
   two sites may have IP interconnectivity over that backbone only if at
   least one of these subsets contains them both.

   The subsets we have created are "Virtual Private Networks" (VPNs).
   Two sites have IP connectivity over the common backbone only if there
   is some VPN which contains them both.  Two sites which have no VPN in
   common have no connectivity over that backbone.

   If all the sites in a VPN are owned by the same enterprise, the VPN
   is a corporate "intranet".  If the various sites in a VPN are owned
   by different enterprises, the VPN is an "extranet".  A site can be in
   more than one VPN; e.g., in an intranet and in several extranets.  In
   general, when we use the term VPN we will not be distinguishing
   between intranets and extranets.

   We wish to consider the case in which the backbone is owned and
   operated by one or more Service Providers (SPs).  The owners of the
   sites are the "customers" of the SPs.  The policies that determine
   whether a particular collection of sites is a VPN are the policies of
   the customers.  Some customers will want the implementation of these
   policies to be entirely the responsibility of the SP.  Other
   customers may want to implement these policies themselves, or to
   share with the SP the responsibility for implementing these policies.
   In this document, we are primarily discussing mechanisms that may be
   used to implement these policies.  The mechanisms we describe are
   general enough to allow these policies to be implemented either by
   the SP alone, or by a VPN customer together with the SP.  Most of the
   discussion is focused on the former case, however.

   The mechanisms discussed in this document allow the implementation of
   a wide range of policies. For example, within a given VPN, we can
   allow every site to have a direct route to every other site ("full
   mesh"), or we can restrict certain pairs of sites from having direct
   routes to each other ("partial mesh").

   In this document, we are interested in the case where the common
   backbone offers an IP service.  We are NOT focused on the case where
   the common backbone is part of the public Internet, but rather on the
   case where it the backbone network of an SP or set of SPs with which
   the customer maintains contractual relationships.  That is, the
   customer is explicitly  purchasing VPN service from the SP, rather



Rosen, et al.                                                   [Page 4]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


   than purchasing Internet access from it.  (The customer may or may
   not be purchasing Internet access from the same SP as well.)

   The customer itself may be a single enterprise, a set of enterprises
   needing an extranet, an Internet Service Provider, an application
   service provider, or even another SP which offers the same kind of
   VPN service to its own customers.

   In the rest of this introduction, we specify some properties which
   VPNs should have.  The remainder of this document outlines a VPN
   model which has all these properties.


1.2. Edge Devices

   We suppose that at each site, there are one or more Customer Edge
   (CE) devices, each of which is "attached" via some sort of data link
   (e.g., PPP, ATM, ethernet, Frame Relay, GRE tunnel, etc.)  to one or
   more Provider Edge (PE) routers.  Routers in the Provider's network
   which do not attach to CE devices are known as "P routers".

   If a particular site has a single host, that host may be the CE
   device.  If a particular site has a single subnet, the CE device may
   be a switch.  In general, the CE device can be expected to be a
   router, which we call the CE router.

   We will say that a PE router is attached to a particular VPN if it is
   attached to a CE device which is in that VPN.  Similarly, we will say
   that a PE router is attached to a particular site if it is attached
   to a CE device which is in that site.

   When the CE device is a router, it is a routing peer of the PE(s) to
   which it is attached, but it is NOT a routing peer of CE routers at
   other sites.  Routers at different sites do not directly exchange
   routing information with each other; in fact, they do not even need
   to know of each other at all.  As a consequence, the customer has no
   backbone or "virtual backbone" to manage, and does not have to deal
   with any inter-site routing issues.  In other words, in the scheme
   described in this document, a VPN is NOT an "overlay" on top of the
   SP's network.

   With respect to the management of the edge devices, clear
   administrative boundaries are maintained between the SP and its
   customers.  Customers are not required to access the PE or P routers
   for management purposes, nor is the SP required to access the CE
   devices for management purposes.





Rosen, et al.                                                   [Page 5]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


1.3. Multiple Forwarding Tables in PEs

   Each PE router maintains a number of separate forwarding tables.
   Every site to which the PE is attached must be mapped to one of those
   forwarding tables.  When a packet is received from a particular site,
   the forwarding table associated with that site is consulted in order
   to determine how to route the packet.  The forwarding table
   associated with a particular site S is populated ONLY with routes
   that lead to other sites which have at least one VPN in common with
   S. This prevents communication between sites which have no VPN in
   common.

   A PE router is attached to a site by virtue of being the endpoint of
   an interface or "sub-interface" (PVC, VLAN, GRE tunnel, etc.) whose
   other endpoint is a CE device.  If there are multiple attachments
   between a site and a PE router, all the attachments may be mapped to
   the same forwarding table, or different attachments may be mapped to
   different forwarding tables.  When a PE router receives a packet from
   a CE device, it knows the interface or sub-interface over which the
   packet arrived, and this determines the forwarding table used for
   processing that packet.  The choice of forwarding table is NOT
   determined by the user content of the packet.

   Different sites can be mapped to the same forwarding table, but ONLY
   if they have all their VPNs in common.


1.4. VPNs with Overlapping Address Spaces

   If two VPNs have no sites in common, then they may have overlapping
   address spaces.  That is, a given address might be used in VPN V1 as
   the address of system S1, but in VPN V2 as the address of a
   completely different system S2.  This is a common situation when the
   VPNs each use an RFC1918 private address space.  (In fact, two VPNs
   which do have sites in common may have overlapping address spaces, as
   long as the overlapping part of the address space does not belong to
   any of the sites which the two VPNs have in common.)

   The fact that sites in different VPNs are mapped to different
   forwarding tables makes it possible for different VPNs to have
   overlapping address spaces, without creating any ambiguity.










Rosen, et al.                                                   [Page 6]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


1.5. VPNs with Different Routes to the Same System

   Although a site may be in multiple VPNs, it is not necessarily the
   case that the route to a given system at that site should be the same
   in all the VPNs.  Suppose, for example, we have an intranet
   consisting of sites A, B, and C, and an extranet consisting of A, B,
   C, and the "foreign" site D.  Suppose that at site A there is a
   server, and we want clients from B, C, or D to be able to use that
   server.  Suppose also that at site B there is a firewall.  We want
   all the traffic from site D to the server to pass through the
   firewall, so that traffic from the extranet can be access controlled.
   However, we don't want traffic from C to pass through the firewall on
   the way to the server, since this is intranet traffic.

   This means that it needs to be possible to set up two routes to the
   server.  One route, used by sites B and C, takes the traffic directly
   to site A.  The second route, used by site D, takes the traffic
   instead to the firewall at site B.  If the firewall allows the
   traffic to pass, it then appears to be traffic coming from site B,
   and follows the route to site A.


1.6. SP Backbone Routers

   The SP's backbone consists of the PE routers, as well as other
   routers ("P routers") which do not attach to CE devices.

   If every router in an SP's backbone had to maintain routing
   information for all the VPNs supported by the SP, this model would
   have severe scalability problems; the number of sites that could be
   supported would be limited by the amount of routing information that
   could be held in a single router.  It is important therefore that the
   routing information about a particular VPN is only required to be
   present in those PE routers which attach to that VPN.  In particular,
   the P routers should not need to have ANY per-VPN routing information
   whatsoever.  (This condition may need to be relaxed somewhat when
   multicast routing is considered.  This is not considered further in
   this paper.)

   So just as the VPN owners do not have a backbone or "virtual
   backbone" to administer, the SPs themselves do not have a separate
   backbone or "virtual backbone" to administer for each VPN.  Site-to-
   site routing in the backbone is optimal (within the constraints of
   the policies used to form the VPNs), and is not constrained in any
   way by an artificial "virtual topology" of tunnels.

   VPNs may span multiple service providers. There are a number of
   possible methods for implementing this, which shall be discussed



Rosen, et al.                                                   [Page 7]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


   later.


1.7. Security

   VPNs of the sort being discussed here, even without making use of
   cryptographic security measures, provide a level of security
   equivalent to that obtainable when a level 2 backbone (e.g., Frame
   Relay) is used.  In the absence of misconfiguration or deliberate
   interconnection of different VPNs, it is not possible for systems in
   one VPN to gain access to systems in another VPN.


2. Sites and CEs

   From the perspective of a particular backbone network, a set of IP
   systems constitutes a site if those systems have mutual IP
   interconnectivity, and communication among them occurs without use of
   the backbone. In general, a site will consist of a set of systems
   which are in geographic proximity.  However, this is not universally
   true.  If two geographic locations are connected via a leased line,
   over which OSPF is running, and if that line is the preferred way of
   communicating between the two locations, then the two locations can
   be regarded as a single site, even if each location has its own CE
   router.  (This notion of "site" is topological, rather than
   geographical.  If the leased line goes down, or otherwise ceases to
   be the preferred route, but the two geographic locations can continue
   to communicate by using the VPN backbone, then one site has become
   two.)

   A CE device is always regarded as being in a single site (though as
   we shall see, a site may consist of multiple "virtual sites"). A
   site, however, may belong to multiple VPNs.

   A PE router may attach to CE devices in any number of different
   sites, whether those CE devices are in the same or in different VPNs.
   A CE device may, for robustness, attach to multiple PE routers, of
   the same or of different service providers.  If the CE device is a
   router, the PE router and the CE router will appear as router
   adjacencies to each other.

   While the basic unit of interconnection is the site, the architecture
   described herein allows a finer degree of granularity in the control
   of interconnectivity. For example, certain systems at a site may be
   members of an intranet as well as members of one or more extranets,
   while other systems at the same site may be restricted to being
   members of the intranet only.




Rosen, et al.                                                   [Page 8]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


   In some cases, a particular site may be divided by the customer into
   several "virtual sites", perhaps by the use of VLANs.  Each virtual
   site may be a member of a different set of VPNs. For example, if a CE
   supports VLANs, and wants each VLAN mapped to a separate VPN, the
   packets sent between CE and PE could be contained in the site's VLAN
   encapsulation.  Then the VLAN tag could be used by the PE, along with
   the interface over which the packet is received, to assign the packet
   to a particular VPN.

   Alternatively, one could divide the interface into multiple "sub-
   interfaces" (particularly if the interface is Frame Relay or ATM),
   and assign the packet to a VPN based on the sub-interface over which
   it arrives.  Or one could simply use a different interface for each
   virtual site.  In any case, only one CE router is ever needed per
   site, even if there are multiple virtual sites.  Of course, a
   different CE router could be used for each virtual site, if that is
   desired.

   Note that in all these cases, the mechanisms, as well as the policy,
   for controlling which traffic is in which VPN are in the hand of the
   customer.

   If it is desired to have a particular host be in multiple virtual
   sites, then that host must determine, for each packet, which virtual
   site the packet is associated with.  It can do this, e.g., by sending
   packets from different virtual sites on different VLANs, our out
   different network interfaces.


3. VRFs: Per-Site Forwarding Tables in the PEs

   Each PE router maintains one or more "per-site forwarding tables."
   These are known as VRFs, or "VPN Routing and Forwarding" tables.
   Every site to which the PE router is attached is associated with one
   of these tables.  A particular packet's IP destination address is
   looked up in a particular VRF only if that packet has arrived
   directly from a site which is associated with that table.

   It would in fact be more precise to say that in the PE router,
     - sub-interfaces may be mapped to VRFs,
     - the mapping is many-to-one,
     - the VRF in which a packet's destination address is looked up is
       determined by the sub-interface over which it is received, and
     - two sub-interfaces may not be mapped to the same VRF unless the
       same set of routes is meant to be available to packets received
       over either sub-interface.

   A sub-interface which is mapped to a VRF may be referred to as a "VRF



Rosen, et al.                                                   [Page 9]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


   sub-interface".

   How are the VRFs populated?

   As an example, let PE1, PE2, and PE3 be three PE routers, and let
   CE1, CE2, and CE3 be three CE routers. Suppose that PE1 learns, from
   CE1, the routes which are reachable at CE1's site.  If PE2 and PE3
   are attached respectively to CE2 and CE3, and there is some VPN V
   containing CE1, CE2, and CE3, then PE1 uses BGP to distribute to PE2
   and PE3 the routes which it has learned from CE1.  PE2 and PE3 use
   these routes to populate the VRFs which they associate respectively
   with the sites of CE2 and CE3.  Routes from sites which are not in
   VPN V do not appear in these VRFs, which means that packets from CE2
   or CE3 cannot be sent to sites which are not in VPN V.

   If a site is in multiple VPNs, the VRF associated with that site
   contains routes from the full set of VPNs of which the site is a
   member.

   A PE generally associates only one VRF to each site, even if it is
   multiply connected to that site.  However, different sites can share
   the same VRF if (and only if) they are meant to use exactly the same
   set of routes.

   When a PE receives a packet from a directly attached site, it always
   looks up the packet's destination address in the VRF which is
   associated with that site.  However, when a PE receives a packet
   which is destined to go to a particular directly attached site, it
   does not necessarily need to lookup the packet's destination address
   in the VRF (or anywhere else).  The packet may already be carrying
   enough information (in the form of an MPLS label, see section 5) to
   determine the packet's outgoing sub-interface.  That is, the packet's
   exit point from the backbone may be completely determined by the
   information in the VRF associated with its entry point to the
   backbone.

   This allows the backbone to support multiple different routes to the
   same system, where the route followed by a given packet is determined
   by the site from which the packet enters the backbone.  E.g., one may
   have one route to a given system for packets from the extranet (where
   the route leads to a firewall), and a different route to the same
   system for packets from the intranet (including packets that have
   already passed through the firewall).








Rosen, et al.                                                  [Page 10]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


4. VPN Route Distribution via BGP

   PE routers use BGP to distribute VPN routes to each other (more
   accurately, to cause VPN routes to be distributed to each other).

   We allow each VPN to have its own address space, which means that a
   given address may denote different systems in different VPNs.  If two
   routes, to the same IP address prefix, are actually routes to
   different systems, it is important to ensure that BGP not treat them
   as comparable.  Otherwise BGP might choose to install only one of
   them, making the other system unreachable.  Further, we must ensure
   that POLICY is used to determine which packets get sent on which
   routes; given that several such routes are installed by BGP, only one
   such must appear in any particular VRF.

   We meet these goals by the use of a new address family, as specified
   below.


4.1. The VPN-IPv4 Address Family

   The BGP Multiprotocol Extensions [BGP-MP] allow BGP to carry routes
   from multiple "address families".  We introduce the notion of the
   "VPN-IPv4 address family".  A VPN-IPv4 address is a 12-byte quantity,
   beginning with an 8-byte "Route Distinguisher (RD)" and ending with a
   4-byte IPv4 address.  If two VPNs use the same IPv4 address prefix,
   the PEs translate these into unique VPN-IPv4 address prefixes.  This
   ensures that if the same address is used in two different VPNs, it is
   possible to install two completely different routes to that address,
   one for each VPN.

   The RD does not by itself impose any semantics; it contains no
   information about the origin of the route or about the set of VPNs to
   which the route is to be distributed.  The purpose of the RD is
   solely to allow one to create distinct routes to a common IPv4
   address prefix.  Other means are used to determine where to
   redistribute the route (see section 4.3).

   The RD can also be used to create multiple different routes to the
   very same system.  In section 3, we gave an example where the route
   to a particular server had to be different for intranet traffic than
   for extranet traffic.  This can be achieved by creating two different
   VPN-IPv4 routes that have the same IPv4 part, but different RDs.
   This allows BGP to install multiple different routes to the same
   system, and allows policy to be used (see section 4.3.5) to decide
   which packets use which route.

   The RDs are structured so that every service provider can administer



Rosen, et al.                                                  [Page 11]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


   its own "numbering space" (i.e., can make its own assignments of
   RDs), without conflicting with the RD assignments made by any other
   service provider.  An RD consists of a two-byte type field, an
   administrator field, and an assigned number field.  The value of the
   type field determines the lengths of the other two fields, as well as
   the semantics of the administrator field.  The administrator field
   identifies an assigned number authority, and the assigned number
   field contains a number which has been assigned, by the identified
   authority, for a particular purpose.  For example, one could have an
   RD whose administrator field contains an Autonomous System number
   (ASN), and whose (4-byte) number field contains a number assigned by
   the SP to whom IANA has assigned that ASN.

   RDs are given this structure in order to ensure that an SP which
   provides VPN backbone service can always create a unique RD when it
   needs to do so. However, the structuring provides no semantics. When
   BGP compares two such address prefixes, it ignores the structure
   entirely.

   Note that VPN-IPv4 addresses and  IPv4 addresses are always
   considered by BGP to be incomparable.

   A VRF may have multiple VPN-IPv4 routes for a single IPv4 address
   prefix.  When a packet's destination address is matched against a
   VPN-IPv4 route, only the IPv4 part is actually matched.

   A PE needs to be configured such that routes which lead to particular
   CE become associated with a particular RD.  The configuration may
   cause all routes leading to the same CE to be associated with the
   same RD, or it may be cause different routes to be associated with
   different RDs, even if they lead to the same CE.


4.2. Encoding of Route Distinguishers

   As stated, a VPN-IPv4 address consists of an 8-byte Route
   Distinguisher followed by a 4-byte IPv4 address.  The RDs are encoded
   as follows:

     - Type Field: 2 bytes
     - Value Field: 6 bytes

   The interpretation of the Value field depends on the value of the
   Type field. At the present time, two values of the type field are
   defined: 0 and 1.






Rosen, et al.                                                  [Page 12]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


     - Type 0: The Value field consists of two subfields:

         * Administrator subfield: 2 bytes
         * Assigned Number subfield: 4 bytes

       The Administrator subfield must contain an Autonomous System
       number. If this ASN is from the public ASN space, it must have
       been assigned by IANA (use of ASN values from the private ASN
       space is strongly discouraged).  The Assigned Number subfield
       contains a number from a numbering space which is administered by
       the enterprise to which the ASN has been assigned by IANA.

     - Type 1: The Value field consists of two subfields:

         * Administrator subfield: 4 bytes
         * Assigned Number subfield: 2 bytes

       The Administrator subfield must contain an IP address. If this IP
       address is from the public IP address space, it must have been
       assigned by IANA (use of addresses from the private IP address
       space is strongly discouraged). The Assigned Number sub-field
       contains a number from a numbering space which is administered by
       the enterprise to which the IP address has been assigned.


4.3. Controlling Route Distribution

   In this section, we discuss the way in which the distribution of the
   VPN-IPv4 routes is controlled.


4.3.1. The Route Target Attribute

   Every VRF is associated with one or more "Route Target" attributes.

   When a VPN-IPv4 route is created by a PE router, it is associated
   with one or more "Route Target" attributes.  These are carried in BGP
   as attributes of the route.

   Any route associated with Route Target T must be distributed to every
   PE router that has a VRF associated with Route Target T.  When such a
   route is received by a PE router, it is eligible to be installed
   those of the PE's VRFs which are associated with Route Target T.
   (Whether it actually gets installed depends on the outcome of the BGP
   decision process.)

   A Route Target attribute can be thought of as identifying a set of
   sites.  (Though it would be more precise to think of it as



Rosen, et al.                                                  [Page 13]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


   identifying a set of VRFs.)  Associating a particular Route Target
   attribute with a route allows that route to be placed in the VRFs
   that are used for routing traffic which is received from the
   corresponding sites.

   There is a set of Route Targets that a PE router attaches to a route
   received from site S; these may be called the "Export Targets". And
   there is a set of Route Targets that a PE router uses to determine
   whether a route received from another PE router could be placed in
   the VRF associated with site S; these may be called the "Import
   Targets". The two sets are distinct, and need not be the same.  Note
   that a particular VPN-IPv4 route is only eligible for installation in
   a particular VRF if there is some Route Target which is both one of
   the route's Route Targets and one of the VRF's Import Targets.

   The function performed by the Route Target attribute is similar to
   that performed by the BGP Communities Attribute.  However, the format
   of the latter is inadequate for present purposes, since it allows
   only a two-byte numbering space.  It is desirable to structure the
   format, similar to what we have described for RDs (see section 4.2),
   so that a type field defines the length of an administrator field,
   and the remainder of the attribute is a number from the specified
   administrator's numbering space.  This can be done using BGP Extended
   Communities.  The Route Targets discussed herein are encoded as BGP
   Extended Community Route Targets [BGP-EXTCOMM].

   When a BGP speaker has received more than one route to the same VPN-
   IPv4 prefix, the BGP rules for route preference are used to choose
   which route are installed.

   Note that a route can only have one RD, but it can have multiple
   Route Targets.  In BGP, scalability is improved if one has a single
   route with multiple attributes, as opposed to multiple routes.  One
   could eliminate the Route Target attribute by creating more routes
   (i.e., using more RDs), but the scaling properties would be less
   favorable.

   How does a PE determine which Route Target attributes to associate
   with a given route?  There are a number of different possible ways.
   The PE might be configured to associate all routes that lead to a
   particular site with a particular Route Target.  Or the PE might be
   configured to associate certain routes leading to a particular site
   with one Route Target, and certain with another.  Or the CE router,
   when it distributes these routes to the PE (see section 7), might
   specify one or more Route Targets for each route.  The latter method
   shifts the control of the mechanisms used to implement the VPN
   policies from the SP to the customer.  If this method is used, it may
   still be desirable to have the PE eliminate any Route Targets that,



Rosen, et al.                                                  [Page 14]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


   according to its own configuration, are not allowed, and/or to add in
   some Route Targets that according to its own configuration are
   mandatory.


4.3.2. Route Distribution Among PEs by BGP

   If two sites of a VPN attach to PEs which are in the same Autonomous
   System, the PEs can distribute VPN-IPv4 routes to each other by means
   of an IBGP connection between them.  Alternatively, each can have an
   IBGP connection to a route reflector.

   When a PE router distributes a VPN-IPv4 route via BGP, it uses its
   own address as the "BGP next hop".  This address is encoded as a
   VPN-IPv4 address with an RD of 0.  ([BGP-MP] requires that the next
   hop address be in the same address family as the NLRI.)  It also
   assigns and distributes an MPLS label.  (Essentially, PE routers
   distribute not VPN-IPv4 routes, but Labeled VPN-IPv4 routes. Cf.
   [MPLS-BGP]).  When the PE processes a received packet that has this
   label at the top of the stack, the PE will pop the stack, and process
   the packet appropriately.

   The PE may distribute the exact set of routes that appears in the
   VRF, or it may perform summarization and distribute aggregates of
   those routes, or it may do some of one and some of the other.

   Suppose that a PE has assigned label L to route R, and has
   distributed this label mapping via BGP.  If R is an aggregate of a
   set of routes in the VRF, the PE will know that packets from the
   backbone which arrive with this label must have their destination
   addresses looked up in a VRF.  When the PE looks up the label in its
   Label Information Base, it learns which VRF must be used.  On the
   other hand, if R is not an aggregate, then when the PE looks up the
   label, it learns the output sub-interface and the data link
   encapsulation header for the packet.  In this case, no lookup in the
   VRF is done.

   We would expect that the most common case would be the case where the
   route is NOT an aggregate.  The case where it is an aggregate can be
   very useful though if the VRF contains a large number of host routes
   (e.g., as in dial-in), or if the VRF has an associated LAN interface
   (where there is a different outgoing layer 2 header for each system
   on the LAN, but a route is not distributed for each such system).
   However, we do not consider this further in this paper.

   Note that the use of BGP-distributed MPLS labels is only possible if
   there is a label switched path between the PE router that installs
   the BGP-distributed route and PE router which is the BGP next hop of



Rosen, et al.                                                  [Page 15]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


   that route.  This label switched path may follow a "best effort"
   route, or it may follow a traffic engineered route.  Between a
   particular PE router and its BGP next hop for a particular route
   there may be one label switched path, or there may be several,
   perhaps with different QoS characteristics.  All that matters for the
   VPN architecture is that some label switched path between the router
   and its BGP next hop exists.  However, to ensure interoperability
   among systems which implement this VPN architecture, all such systems
   must support LDP [MPLS-LDP].

   A PE router, UNLESS it is a Route Reflector (see section 4.3.3)
   should not install a VPN-IPv4 route unless it has at least one VRF
   with an Import Target identical to one of the route's Route Target
   attributes.  Inbound filtering should be used to cause such routes to
   be discarded.  If a new Import Target is later added to one of the
   PE's VRFs (a "VPN Join" operation), it must then acquire the routes
   it may previously have discarded.  This can be done using the refresh
   mechanism described in [BGP-RFSH].  The outbound route filtering
   mechanism of [BGP-ORF] can also be used to advantage to make the
   filtering more dynamic.

   Similarly, if a particular Import Target is no longer present in any
   of a PE's VRFs (as a result of one or more "VPN Prune" operations),
   the PE may discard all routes which, as a result, no longer have any
   of the PE's VRF's Import Targets as one of their Route Target
   Attributes.

   A router which is not attached to any VPN, and which is not a Route
   Reflector (i.e., a P router), never installs any VPN-IPv4 routes at
   all.

   Note that VPN Join and Prune operations are non-disruptive, and do
   not require any BGP connections to be brought down.

   As a result of these distribution rules, no one PE ever needs to
   maintain all routes for all VPNs; this is an important scalability
   consideration.


4.3.3. Use of Route Reflectors

   Rather than having a complete IBGP mesh among the PEs, it is
   advantageous to make use of BGP Route Reflectors [BGP-RR] to improve
   scalability.  All the usual techniques for using route reflectors to
   improve scalability, e.g., route reflector hierarchies, are
   available.

   Route reflectors are the only systems which need to have routing



Rosen, et al.                                                  [Page 16]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


   information for VPNs to which they are not directly attached.
   However, there is no need to have any one route reflector know all
   the VPN-IPv4 routes for all the VPNs supported by the backbone.

   We outline below two different ways to partition the set of VPN-IPv4
   routes among a set of route reflectors.

      1. Each route reflector is preconfigured with a list of Route
         Targets. For redundancy, more than one route reflector may be
         preconfigured with the same list. A route reflector uses the
         preconfigured list of Route Targets to construct its inbound
         route filtering.  The route reflector may use the techniques of
         [BGP-ORF] to install on each of its peers (regardless of
         whether the peer is another route reflector, or a PE) the set
         of "Outbound Route Filters" (ORFs) that contain the list of its
         preconfigured Route Targets. Note that route reflectors should
         accept ORFs from other route reflectors, which means that route
         reflectors should advertise the ORF capability to other route
         reflectors.

         A service provider may modify the list of preconfigured Route
         Targets on a route reflector. When this is done, the route
         reflector modifies the ORFs it installs on all of its IBGP
         peers. To reduce the frequency of configuration changes on
         route reflectors, each route reflector may be preconfigured
         with a block of Route Targets.  This way, when a new Route
         Target is needed for a new VPN, there is already one or more
         route reflectors that are (pre)configured with this Route
         Target.

         Unless a given PE is a client of all route reflectors, when a
         new VPN is added to the PE ("VPN Join"), it will need to become
         a client of the route reflector(s) that maintain routes for
         that VPN. Likewise, deleting an existing VPN from the PE ("VPN
         Prune") may result in a situation where the PE no longer need
         to be a client of some route reflector(s).  In either case, the
         Join or Prune operation is non-disruptive, and never requires a
         BGP connection to be brought down, only to be brought right
         back up.

         (By "adding a new VPN to a PE", we really mean adding a new
         import Route Target to one of its VRFs, or adding a new VRF
         with an import Route Target not had by any of the PE's other
         VRFs.)







Rosen, et al.                                                  [Page 17]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


      2. Another method is to have each PE be a client of some subset of
         the route reflectors. A route reflector is not preconfigured
         with the list of Route Targets, and does not perform inbound
         route filtering of routes received from its clients (PEs);
         rather it accepts all the routes received from all of its
         clients (PEs).  The route reflector keeps track of the set of
         the Route Targets carried by all the routes it receives.  When
         the route reflector receives from its client a route with a
         Route Target that is not in this set, this Route Target is
         immediately added to the set. On the other hand, when the route
         reflector no longer has any routes with a particular Route
         Target that is in the set, the route reflector should delay (by
         a few hours) the deletion of this Route Target from the set.

         The route reflector uses this set to form the inbound route
         filters that it applies to routes received from other route
         reflectors. The route reflector may also use ORFs to install
         the appropriate outbound route filtering on other route
         reflectors. Just like with the first approach, a route
         reflector should accept ORFs from other route reflectors. To
         accomplish this, a route reflector advertises ORF capability to
         other route reflectors.

         When the route reflector changes the set, it should immediately
         change its inbound route filtering. In addition, if the route
         reflector uses ORFs, then the ORFs have to be immediately
         changed to reflect the changes in the set. If the route
         reflector doesn't use ORFs, and a new Route Target is added to
         the set, the route reflector, after changing its inbound route
         filtering, must issue BGP Refresh to other router reflectors.

         With this procedure, VPN Join and Prune operations are also
         non-disruptive.

   Just as there is no one PE router that needs to know all the VPN-IPv4
   routes that are supported over the backbone, these distribution rules
   ensure that there is no one RR which needs to know all the VPN-IPv4
   routes that are supported over the backbone.  As a result, the total
   number of such routes that can be supported over the backbone is not
   bounded by the capacity of any single device, and therefore can
   increase virtually without bound.










Rosen, et al.                                                  [Page 18]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


4.3.4. How VPN-IPv4 NLRI is Carried in BGP

   The BGP Multiprotocol Extensions [BGP-MP] are used to encode the
   NLRI.  If the AFI field is set to 1, and the SAFI field is set to
   128, the NLRI is an MPLS-labeled VPN-IPv4 address.  AFI 1 is used
   since the network layer protocol associated with the NLRI is still
   IP.  Note that this VPN architecture does not require the capability
   to distribute unlabeled VPN-IPv4 addresses.

   In order for two BGP speakers to exchange labeled VPN-IPv4 NLRI, they
   must use BGP Capabilities Negotiation to ensure that they both are
   capable of properly processing such NLRI.  This is done as specified
   in [BGP-MP], by using capability code 1 (multiprotocol BGP), with an
   AFI of 1 and an SAFI of 128.

   The labeled VPN-IPv4 NLRI itself is encoded as specified in [MPLS-
   BGP], where the prefix consists of an 8-byte RD followed by an IPv4
   prefix.


4.3.5. Building VPNs using Route Targets

   By setting up the Import Targets and Export Targets properly, one can
   construct different kinds of VPNs.

   Suppose it is desired to create a a fully meshed closed user group,
   i.e., a set of sites where each can send traffic directly to the
   other, but traffic cannot be sent to or received from other sites.
   Then each site is associated with a VRF, a single Route Target
   attribute is chosen, that Route Target is assigned to each VRF as
   both the Import Target and the Export Target, and that Route Target
   is not assigned to any other VRFs as either the Import Target or the
   Export Target.

   Alternatively, suppose one desired, for whatever reason, to create a
   "hub and spoke" kind of VPN.  This could be done by the use of two
   Route Target values, one meaning "Hub" and one meaning "Spoke".  At
   the VRFs attached to the hub sites, "Hub" is the Export Target and
   "Spoke" is the Import Target.  At the VRFs attached to the spoke
   site, "Hub" is the Import Target and "Spoke" is the Export Target.

   Thus the methods for controlling the distribution of routing
   information among various sets of sites are very flexible, which in
   turn provides great flexibility in constructing VPNs.







Rosen, et al.                                                  [Page 19]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


5. Forwarding Across the Backbone

   If the intermediate routers in the backbone do not have any
   information about the routes to the VPNs, how are packets forwarded
   from one VPN site to another?

   This is done by means of MPLS with a two-level label stack.

   PE routers (and ASBRs which redistribute VPN-IPv4 addresses) need to
   insert /32 address prefixes for themselves into the IGP routing
   tables of the backbone.  This enables MPLS, at each node in the
   backbone network, to assign a label corresponding to the route to
   each PE router.  To ensure interoperability among different
   implementations, it is required to support LDP for setting up the
   label switched paths across the backbone.  However, other methods of
   setting up these label switched paths are also possible.  (Some of
   these other methods may not require the presence of the /32 address
   prefixes in the IGP.)

   When a PE receives a packet from a CE device, it chooses a particular
   VRF in which to look up the packet's destination address.  This
   choice is based on the packet's incoming sub-interface.  Assume that
   a match is found.  As a result we learn a "next hop" and an "outgoing
   sub-interface".

   If the packet's outgoing sub-interface is associated with a VRF, then
   the next hop is a CE device.  The packet is sent directly to the CE
   device. However, if the outgoing sub-interface and the incoming sub-
   interface are associated with different VRFs, and the route which
   best matches the destination address in the incoming sub-interface's
   VRF is an aggregate of several routes in the outgoing sub-interface's
   VRF, it may be necessary to look up the packet's destination address
   in the VRF of the outgoing interface as well.

   If the packet's outgoing sub-interface is NOT associated with a VRF,
   then the packet must travel at least one hop through the backbone.
   The packet thus has a "BGP Next Hop", and the BGP Next Hop will have
   assigned a label for the route which best matches the packet's
   destination address.  This label is pushed onto the packet's label
   stack, and becomes the bottom label.  The packet will also have an
   "IGP Next Hop", which is the next hop along the IGP route to the BGP
   Next Hop.  The IGP Next Hop will have assigned a label for the route
   which best matches the address of the BGP Next Hop.  This label gets
   pushed on as the packet's top label.  The packet is then forwarded to
   the IGP next hop.  (Of course, if the BGP Next Hop and the IGP Next
   Hop are the same, and if penultimate hop popping is used, the packet
   may be sent with only the BGP-supplied label.)




Rosen, et al.                                                  [Page 20]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


   MPLS will then carry the packet across the backbone.  The egress PE
   router's treatment of the packet will depend on the label that was
   first pushed on by the ingress PE.  In many cases, the PE will be
   able to determine, from this label, the sub-interface over which the
   packet should be transmitted (to a CE device), as well as the proper
   data link layer header for that interface.  In other cases, the PE
   may only be able to determine that the packet's destination address
   needs to be looked up in a particular VRF before being forwarded to a
   CE device.  Information in the MPLS header itself, and/or information
   associated with the label, may also be used to provide QoS on the
   interface to the CE.  In any event, when the packet finally gets to a
   CE device, it will again be an ordinary unlabeled IP packet.

   Note that it is the two-level labeling that makes it possible to keep
   all the VPN routes out of the P routers, and this in turn is crucial
   to ensuring the scalability of the model.  The backbone does not even
   need to have routes to the CEs, only to the PEs.


6. Maintaining Proper Isolation of VPNs

   To maintain proper isolation of one VPN from another, it is important
   that no router in the backbone accept a labeled packet from any
   adjacent non-backbone device unless the following two conditions
   hold:

      1. the label at the top of the label stack was actually
         distributed by that backbone router to that non-backbone
         device, and

      2. the backbone router can determine that use of that label will
         cause the packet to leave the backbone before any labels lower
         in the stack will be inspected, and before the IP header will
         be inspected.

   The first condition ensure that any labeled packets received from
   non-backbone routers have a legitimate and properly assigned label at
   the top of the label stack.  The second condition ensures that the
   backbone routers will never look below that top label.  Of course,
   the simplest way to meet these two conditions is just to have the
   backbone devices refuse to accept labeled packets from non-backbone
   devices.









Rosen, et al.                                                  [Page 21]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


7. How PEs Learn Routes from CEs

   The PE routers which attach to a particular VPN need to know, for
   each of that VPN's sites, which addresses in that VPN are at each
   site.

   In the case where the CE device is a host or a switch, this set of
   addresses will generally be configured into the PE router attaching
   to that device.  In the case where the CE device is a router, there
   are a number of possible ways that a PE router can obtain this set of
   addresses.

   The PE translates these addresses into VPN-IPv4 addresses, using a
   configured RD.  The PE then treats these VPN-IPv4 routes as input to
   BGP.  Routes from a site are not leaked into the backbone's IGP.

   Exactly which PE/CE route distribution techniques are possible
   depends on whether a particular CE is in a "transit VPN" or not.  A
   "transit VPN" is one which contains a router that receives routes
   from a "third party" (i.e., from a router which is not in the VPN,
   but is not a PE router), and that redistributes those routes to a PE
   router.  A VPN which is not a transit VPN is a "stub VPN".  The vast
   majority of VPNs, including just about all corporate enterprise
   networks, would be expected to be "stubs" in this sense.

   The possible PE/CE distribution techniques are:

      1. Static routing (i.e., configuration) may be used. (This is
         likely to be useful only in stub VPNs.)

      2. PE and CE routers may be RIP peers, and the CE may use RIP to
         tell the PE router the set of address prefixes which are
         reachable at the CE router's site.  When RIP is configured in
         the CE, care must be taken to ensure that address prefixes from
         other sites (i.e., address prefixes learned by the CE router
         from the PE router) are never advertised to the PE.  More
         precisely:  if a PE router, say PE1, receives a VPN-IPv4 route
         R1, and as a result distributes an IPv4 route R2 to a CE, then
         R2 must not be distributed back from that CE's site to a PE
         router, say PE2, (where PE1 and PE2 may be the same router or
         different routers), unless PE2 maps R2 to a VPN-IPv4 route
         which is different than (i.e., contains a different RD than)
         R1.

      3. The PE and CE routers may be OSPF peers.  A PE router which is
         an OSPF peer of a CE router appears, to the CE router, to be an
         area 0 router.  If a PE router is an OSPF peer of CE routers
         which are in distinct VPNs, the PE must of course be running



Rosen, et al.                                                  [Page 22]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


         multiple instances of OSPF.

         IPv4 routes which the PE learns from the CE via OSPF are
         redistributed into BGP as VPN-IPv4 routes.  Extended community
         attributes are used to carry, along with the route, all the
         information needed to enable the route to be distributed to
         other CE routers in the VPN in the proper type of OSPF LSA.
         OSPF route tagging is used to ensure that routes received from
         the MPLS/BGP backbone are not sent back into the backbone.

         Specification of the complete set of procedures for the use of
         OSPF between PE and CE must be left to another document.

      4. The PE and CE routers may be BGP peers, and the CE router may
         use BGP (in particular, EBGP to tell the PE router the set of
         address prefixes which are at the CE router's site. (This
         technique can be used in stub VPNs or transit VPNs.)

         From a purely technical perspective, this is by far the best
         technique:

            a) Unlike the IGP alternatives, this does not require the PE
               to run multiple routing algorithm instances in order to
               talk to multiple CEs

            b) BGP is explicitly designed for just this function:
               passing routing information between systems run by
               different administrations

            c) If the site contains "BGP backdoors", i.e., routers with
               BGP connections to routers other than PE routers, this
               procedure will work correctly in all circumstances.  The
               other procedures may or may not work, depending on the
               precise circumstances.

            d) Use of BGP makes it easy for the CE to pass attributes of
               the routes to the PE.  A complete specification of the
               set of attributes and their use is outside the scope of
               this document.  However, some examples of the way this
               may be used are the following:

                 - The CE may suggest a particular Route Target for each
                   route, from among the Route Targets that the PE is
                   authorized to attach to the route.  The PE would then
                   attach only the suggested Route Target, rather than
                   the full set.   This gives the CE  administrator some
                   dynamic control of the distribution of routes from
                   the CE.



Rosen, et al.                                                  [Page 23]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


                 - Additional types of Extended Community attributes may
                   be defined, where the intention is to have those
                   attributes passed transparently from CE to CE.  This
                   would allow CE administrators to implement additional
                   route filtering, beyond that which is done by the
                   PEs.  This additional filtering would not require
                   coordination with the SP.

         On the other hand, using BGP is likely to be something new for
         the CE administrators, except in the case where the customer
         itself is already an Internet Service Provider (ISP), or where
         the CE devices are managed by the SP.

         If a site is not in a transit VPN, note that it need not have a
         unique Autonomous System Number (ASN).  Every CE whose site
         which is not in a transit VPN can use the same ASN.  This can
         be chosen from the private ASN space, and it will be stripped
         out by the PE.  Routing loops are prevented by use of the Site
         of Origin Attribute (see below).

         What if a set of sites constitute a transit VPN?  This will
         generally be the case only if the VPN is itself an ISP's
         network, where the ISP is itself buying backbone services from
         another SP.  The latter SP may be called a "Carrier's Carrier".
         In this case, the best way to provide the VPN is to have the CE
         routers support MPLS, and to use the technique described in
         section 9.


   When we do not need to distinguish among the different ways in which
   a PE can be informed of the address prefixes which exist at a given
   site, we will simply say that the PE has "learned" the routes from
   that site.

   Before a PE can redistribute a VPN-IPv4 route learned from a site, it
   must assign a Route Target attribute (see section 4.3.1) to the
   route, and it may assign a Site of Origin attribute to the route.

   The Site of Origin attribute, if used, is encoded as a Route Origin
   Extended Community [BGP-EXTCOMM].  The purpose of this attribute is
   to uniquely identify the set of routes learned from a particular
   site.  This attribute is needed in some cases to ensure that a route
   learned from a particular site via a particular PE/CE connection is
   not distributed back to the site through a different PE/CE
   connection.  It is particularly useful if BGP is being used as the
   PE/CE protocol, but different sites have not been assigned distinct
   ASNs.




Rosen, et al.                                                  [Page 24]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


8. How CEs learn Routes from PEs

   In this section, we assume that the CE device is a router.

   If the PE places a particular route in the VRF is uses to route
   packets received from a particular CE, then in general, the PE may
   distribute that route to the CE.  Of course the PE may distribute
   that route to the CE only if this is permitted by the rules of the
   PE/CE protocol.  (For example, if a particular PE/CE protocol has
   "split horizon", certain routes in the VRF cannot be redistributed
   back to the CE.)  We add one more restriction on the distribution of
   routes from PE to CE: if a route's Site of Origin attribute
   identifies a particular site, that route must never be redistributed
   to any CE at that site.

   In most cases, however, it will be sufficient for the PE to simply
   distribute the default route to the CE.  (In some cases, it may even
   be sufficient for the CE to be configured with a default route
   pointing to the PE.)  This will generally work at any site which does
   not itself need to distribute the default route to other sites.
   (E.g., if one site in a corporate VPN has the corporation's access to
   the Internet, that site might need to have default distributed to the
   other site, but one could not distribute default to that site
   itself.)

   Whatever procedure is used to distribute routes from CE to PE will
   also be used to distribute routes from PE to CE.


9. Carriers' Carriers

   Sometimes a VPN may actually be the network of an ISP, with its own
   peering and routing policies.  Sometimes a VPN may be the network of
   an SP which is offering VPN services in turn to its own customers.
   VPNs like these can also obtain backbone service from another SP, the
   "carrier's carrier", using essentially the same methods described in
   this document.  In particular:

     - The CE routers should distribute to the PE routers ONLY those
       routes which are internal to the VPN.  This allows the VPN to be
       handled as a stub VPN.

     - The CE routers should support MPLS, in that they should be able
       to receive labels from the PE routers,  and send labeled packets
       to the PE routers.  They do not need to distribute labels of
       their own though.





Rosen, et al.                                                  [Page 25]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


     - The PE routers should distribute, to the CE routers, labels for
       the routes they distribute to the CE routers.

     - Routers at the different sites should establish BGP connections
       among themselves for the purpose of exchanging external routes.

     - All the external routes must be known to the CE routers.

   Then when a CE router looks up a packet's destination address, the
   routing lookup will resolve to an internal address, usually the
   address of the packet's BGP next hop.  The CE labels the packet
   appropriately and sends the packet to the PE.

   In the above procedure, the CE routers are the only routers in the
   VPN which need to support MPLS.  If, on the other hand, all the
   routers at a particular VPN site support MPLS, then it is no longer
   required that the CE routers know all the external routes.  All that
   is required is that the external routes be known to whatever routers
   are responsible for putting the label stack on a hitherto unlabeled
   packet, and that there be label switched path that leads from those
   routers to their BGP peers at other sites.  In this case, for each
   internal route that a CE router distributes to a PE router, it must
   also distribute a label.


10. Inter-Provider Backbones

   What if two sites of a VPN are connected to different Autonomous
   Systems (e.g., because the sites are connected to different SPs)?
   The PE routers attached to that VPN will then not be able to maintain
   IBGP connections with each other, or with a common route reflector.
   Rather, there needs to be some way to use EBGP to distribute VPN-IPv4
   addresses.

   There are a number of different ways of handling this case, which we
   present in order of increasing scalability.

      a) VRF-to-VRF connections at the AS border routers.

         In this procedure, a PE router in one AS attaches directly to a
         PE router in another.  The two PE routers will be attached by
         multiple sub-interfaces, at least one for each of the VPNs
         whose routes need to be passed from AS to AS.  Each PE will
         treat the other as if it were a CE router.  That is, the PEs
         associate each such sub-interface with a VRF, and use EBGP to
         distribute unlabeled IPv4 addresses to each other.

         This is a procedure that "just works", and that does not



Rosen, et al.                                                  [Page 26]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


         require MPLS at the border between ASes.  However, it does not
         scale as well as the other procedures discussed below.

      b) EBGP redistribution of labeled VPN-IPv4 routes from AS to
         neighboring AS.

         In this procedure, the PE routers use IBGP to redistribute
         labeled VPN-IPv4 routes either to an Autonomous System Border
         Router (ASBR), or to a route reflector of which an ASBR is a
         client.  The ASBR then uses EBGP to redistribute those labeled
         VPN-IPv4 routes to an ASBR in another AS, which in turn
         distributes them to the PE routers in that AS, or perhaps to
         another ASBR which in turn distributes them ...

         When using this procedure, VPN-IPv4 routes should only be
         accepted on EBGP connections at private peering points, as part
         of a trusted arrangement between SPs.  VPN-IPv4 routes should
         neither be distributed to nor accepted from the public
         Internet, or from any BGP peers which are not trusted.  An ASBR
         should never accept a labeled packet from an EBGP peer unless
         it has actually distributed the top label to that peer.

         If there are many VPNs having sites attached to different
         Autonomous Systems, there does not need to be a single ASBR
         between those two ASes which holds all the routes for all the
         VPNs; there can be multiple ASBRs, each of which holds only the
         routes for a particular subset of the VPNs.

         This procedure requires that there be a label switched path
         leading from a packet's ingress PE to its egress PE.  Hence the
         appropriate trust relationships must exist between and among
         the set of ASes along the path.  Also, there must be agreement
         among the set of SPs as to which border routers need to receive
         routes with which Route Targets.

      c) Multihop EBGP redistribution of labeled VPN-IPv4 routes between
         source and destination ASes, with EBGP redistribution of
         labeled IPv4 routes from AS to neighboring AS.

         In this procedure, VPN-IPv4 routes are neither maintained nor
         distributed by the ASBRs.  However, an ASBR does use EBGP to
         distribute labeled IPv4 /32 routes to the PE routers within its
         AS.  ASBRs in any transit ASes will also have to use EBGP to
         pass along the labeled /32 routes.  This results in the
         creation of a label switched path from the ingress PE router to
         the egress PE router.  Now PE routers in different ASes can
         establish multi-hop EBGP connections to each other, and can
         exchange VPN-IPv4 routes over those connections.



Rosen, et al.                                                  [Page 27]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


         If the /32 routes for the PE routers are made known to the P
         routers of each AS, everything works normally.  If the /32
         routes for the PE routers are NOT made known to the P routers
         (other than the ASBRs), then this procedure requires a packet's
         ingress PE to put a three label stack on it.  The bottom label
         is assigned by the egress PE, corresponding to the packet's
         destination address in a particular VRF.  The middle label is
         assigned by the ASBR, corresponding to the /32 route to the
         egress PE.  The top label is assigned by the ingress PE's IGP
         Next Hop, corresponding to the /32 route to the ASBR.

         To improve scalability, one can have the multi-hop EBGP
         connections exist only between a route reflector in one AS and
         a route reflector in another.  (However, when the route
         reflectors distribute routes over this connection, they do not
         modify the BGP next hop attribute of the routes.)  The actual
         PE routers would then only have IBGP connections to the route
         reflectors in their own AS.

         This procedure is very similar to the "Carrier's Carrier"
         procedures described in section 9. Like the previous procedure,
         it requires that there be a label switched path leading from a
         packet's ingress PE to its egress PE.


11. Accessing the Internet from a VPN

   Many VPN sites will need to be able to access the public Internet, as
   well as to access other VPN sites.  There are a number of ways to do
   this.

      1. In some VPNs, one or more of the sites will obtain Internet
         Access by means of an "Internet gateway" (perhaps a firewall)
         attached to a non-VRF interface to an ISP.  The ISP may or may
         not be the same organization as the SP which is providing the
         VPN service.

         In this case, the sites which have Internet Access may be
         distributing a default route to their PEs, which in turn
         redistribute it to other PEs and hence into other sites of the
         VPN.  This provides Internet Access for all of the VPN's sites.

         In order to properly handle traffic from the Internet, the ISP
         must distribute, to the Internet, routes leading to addresses
         that are within the VPN.  This is completely independent of any
         of the route distribution procedures described in this
         document.  The internal structure of the VPN will in general
         not be visible from the Internet.



Rosen, et al.                                                  [Page 28]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


         In this model, there is no exchange of routes between a PE
         router's Internet forwarding table and any of its VRFs.  VPN
         route distribution procedures and Internet route distribution
         procedures are completely independent.

         Note that although some sites of the VPN use a VRF interface to
         communicate with the Internet, ultimately all packets to/from
         the Internet traverse a non-VRF interface before
         leaving/entering the VPN, so we refer to this as "non-VRF
         Internet Access".

      2. Some VPNs may obtain Internet access via a VRF interface ("VRF
         Internet Access").  If a packet is received by a PE over a VRF
         interface, and if the packet's destination address does not
         match any route in the VRF, then it may be matched against the
         PE's Internet forwarding table.  If a match is made there, the
         packet can be forwarded natively through the backbone to the
         Internet, instead of being forwarded by MPLS.

         In order for traffic to flow natively in the opposite direction
         (from Internet to VRF interface), some of the routes from the
         VRF must be exported to the Internet forwarding table.
         Needless to say, any such routes must correspond to globally
         unique addresses.

      3. Suppose the PE has the capability to store "non-VPN routes" in
         a VRF.  If a packet's destination address matches a "non-VPN
         route", then the packet is transmitted natively, rather than
         being transmitted via MPLS.  If the VRF contains a non-VPN
         default route, all packets for the public Internet will match
         it, and be forwarded natively to the default route's next hop.
         At that next hop, the packets' destination addresses will be
         lookup up in the Internet forwarding table, and may match more
         specific routes.

         This technique would only be available if none of the CE
         routers is distributing a default route.

      4. It is also possible to obtain Internet access via a VRF
         interface by having the VRF contain the Internet routes.
         Compared with model 2, this eliminates the second lookup, but
         it has the disadvantage of requiring the Internet routes to be
         replicated in each such VRF.

         If this technique is used, the SP may want to make its
         interface to the Internet be a VRF interface, and to use the
         techniques of section 4 to distribute Internet routes, as VPN-
         IPv4 routes, to other VRFs.



Rosen, et al.                                                  [Page 29]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


   It should be clearly understood that by default, there is no exchange
   of routes between a VRF and an Internet forwarding table.  This is
   done ONLY upon agreement between a customer and a SP, and only if it
   suits the customer's policies.


12. Management VPNs

   This specification does not require that the sub-interface connecting
   a PE router and a CE router be a "numbered" interface.  If it is a
   numbered interface, this specification allows the addresses assigned
   to the interface to come from either the address space of the VPN or
   the address space of the SP.

   If a CE router is being managed by the Service Provider, then the
   Service Provider will likely have a network management system which
   needs to be able to communicate with the CE router.  In this case,
   the addresses assigned to the sub-interface connecting the CE and PE
   routers should come from the SP's address space, and should be unique
   within that space.  The network management system should itself
   connect to a PE router (more precisely, be at a site which connects
   to a PE router) via a VRF interface.  The address of the network
   management system will be exported to all VRFs which are associated
   with interfaces to CE routers that are managed by the SP.  The
   addresses of the CE routers will be exported to the VRF associated
   with the Network Management system, but not to any other VRFs.

   This allows communication between CE and Network Management system,
   but does not allow any undesired communication to or among the CE
   routers.

   One way to ensure that the proper route import/exports are done is to
   use two Route Targets, call them T1 and T2.  If a particular VRF
   interface attaches to a CE router that is managed by the SP, then
   that VRF is configured to:

     - import routes that have T1 attached to them, and

     - attach T2 to addresses assigned to each end of its VRF
       interfaces.

   If a particular VRF interface attaches to the SP's Network Management
   system, then that VRF is configured to attach T1 to the address of
   that system, and to import routes that have T2 attached to them.







Rosen, et al.                                                  [Page 30]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


13. Security

   Under the following conditions:

      1. labeled packets are not accepted by backbone routers from
         untrusted or unreliable sources, unless it is known that such
         packets will leave the backbone before the IP header or any
         labels lower in the stack will be inspected, and

      2. labeled VPN-IPv4 routes are not accepted from untrusted or
         unreliable sources,

   the security provided by this architecture is virtually identical to
   that provided to VPNs by Frame Relay or ATM backbones.

   It is worth noting that the use of MPLS makes it much simpler to
   provide this level of security than would be possible if one
   attempted to use some form of IP tunneling in place of the MPLS outer
   label.  It is a simple matter to refuse to accept a labeled packet
   unless the first of the above conditions applies to it.  It is rather
   more difficult to configure a router to refuse to accept an IP packet
   if that packet is an IP tunnelled packet which is going to a "wrong"
   place.




14. Quality of Service

   Although not the focus of this paper, Quality of Service is a key
   component of any VPN service.  In MPLS/BGP VPNs, existing L3 QoS
   capabilities can be applied to labeled packets through the use of the
   "experimental" bits in the shim header [MPLS-ENCAPS], or, where ATM
   is used as the backbone, through the use of ATM QoS capabilities.
   The traffic engineering work discussed in [MPLS-RSVP] is also
   directly applicable to MPLS/BGP VPNs.  Traffic engineering could even
   be used to establish label switched paths with particular QoS
   characteristics between particular pairs of sites, if that is
   desirable.  Where an MPLS/BGP VPN spans multiple SPs, the
   architecture described in [PASTE] may be useful.  An SP may apply
   either intserv or diffserv capabilities to a particular VPN, as
   appropriate.









Rosen, et al.                                                  [Page 31]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


15. Scalability

   We have discussed scalability issues throughout this paper.  In this
   section, we briefly summarize the main characteristics of our model
   with respect to scalability.

   The Service Provider backbone network consists of (a) PE routers, (b)
   BGP Route Reflectors, (c) P routers (which are neither PE routers nor
   Route Reflectors), and, in the case of multi-provider VPNs, (d)
   ASBRs.

   P routers do not maintain any VPN routes.  In order to properly
   forward VPN traffic, the P routers need only maintain routes to the
   PE routers and the ASBRs. The use of two levels of labeling is what
   makes it possible to keep the VPN routes out of the P routers.

   A PE router maintains VPN routes, but only for those VPNs to which it
   is directly attached.

   Route reflectors can be partitioned among VPNs so that each partition
   carries routes for only a subset of the VPNs supported by the Service
   Provider.  Thus no single route reflector is required to maintain
   routes for all VPNs.

   For inter-provider VPNs, if the ASBRs maintain and distribute VPN-
   IPv4 routes, then the ASBRs can be partitioned among VPNs in a
   similar manner, with the result that no single ASBR is required to
   maintain routes for all the inter-provider VPNs.  If multi-hop EBGP
   is used, then the ASBRs need not maintain and distribute VPN-IPv4
   routes at all.

   As a result, no single component within the Service Provider network
   has to maintain all the routes for all the VPNs.  So the total
   capacity of the network to support increasing numbers of VPNs is not
   limited by the capacity of any individual component.


16. Intellectual Property Considerations

   Cisco Systems may seek patent or other intellectual property
   protection for some of all of the technologies disclosed in this
   document. If any standards arising from this document are or become
   protected by one or more patents assigned to Cisco Systems, Cisco
   intends to disclose those patents and license them on reasonable and
   non-discriminatory terms.






Rosen, et al.                                                  [Page 32]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


17. Acknowledgments

   Significant contributions to this work have been made by Ravi
   Chandra, Dan Tappan and Bob Thomas.

   We also wish to thank Shantam Biswas for his review and
   contributions.


18. Authors' Addresses


      Eric C. Rosen
      Cisco Systems, Inc.
      250 Apollo Drive
      Chelmsford, MA, 01824
      E-mail: erosen@cisco.com


      Yakov Rekhter
      Cisco Systems, Inc.
      170 Tasman Drive
      San Jose, CA, 95134
      E-mail: yakov@cisco.com


      Tony Bogovic
      Telcordia Technologies
      445 South Street, Room 1A264B
      Morristown, NJ 07960
      E-mail: tjb@research.telcordia.com


      Stephen John Brannon
      Swisscom AG
      Postfach 1570
      CH-8301
      Glattzentrum (Zuerich), Switzerland
      E-mail: stephen.brannon@swisscom.com












Rosen, et al.                                                  [Page 33]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000



      Marco Carugi
      France Telecom / CNET Research Centre
      IP networks and services
      CNET/DAC/NTR
      Technopole Anticipa
      2, av. P. Marzin
      22307 Lannion
      E-mail: marco.carugi@cnet.francetelecom.fr


      Christopher J. Chase
      AT&T
      200 Laurel Ave
      Middletown, NJ 07748
      USA
      E-mail: chase@att.com


      Ting Wo Chung
      Bell Nexxia
      181 Bay Street
      Suite 350
      Toronto, Ontario
      M5J2T3
      E-mail: ting_wo.chung@bellnexxia.com


      Eric Dean
      Global One
      12490 Sunrise Valley Dr.
      Reston, VA 20170 USA
      E-mail: edean@gip.net


      Jeremy De Clercq
      Alcatel Network Strategy Group
      Francis Wellesplein 1
      2018 Antwerp, Belgium
      E-mail: jeremy.declercq@alcatel.be











Rosen, et al.                                                  [Page 34]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000



      Paul Hitchin
      BT
      BT Adastral Park
      Martlesham Heath,
      Ipswich IP5 3RE
      UK
      E-mail: paul.hitchen@bt.com


      Manoj Leelanivas
      Juniper Networks, Inc.
      385 Ravendale Drive
      Mountain View, CA 94043 USA
      E-mail: manoj@juniper.net


      Dave Marshall
      Worldcom
      901 International Parkway
      Richardson, Texas 75081
      E-mail: dave.marshall@wcom.com


      Luca Martini
      Level 3 Communications, LLC.
      1025 Eldorado Blvd.
      Broomfield, CO, 80021
      E-mail: luca@level3.net


      Monique Jeanne Morrow
      Swisscom AG
      Postfach 1570
      CH-8301
      Glattzentrum (Zuerich), Switzerland
      E-mail: monique.morrow@swisscom.com


      Ravichander Vaidyanathan
      Telcordia Technologies
      445 South Street, Room 1C258B
      Morristown, NJ 07960
      E-mail: vravi@research.telcordia.com







Rosen, et al.                                                  [Page 35]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000



      Adrian Smith
      BT
      BT Adastral Park
      Martlesham Heath,
      Ipswich IP5 3RE
      UK
      E-mail: adrian.ca.smith@bt.com


      Vijay Srinivasan
      1200 Bridge Parkway
      Redwood City, CA 94065
      E-mail: vsriniva@cosinecom.com


      Alain Vedrenne
      SITA EQUANT
      3100 Cumberland Blvd, Suite 200
      Atlanta, GA, 30339 USA
      Email:Alain.Vedrenne@sita.int
            Alain.Vedrenne@equant.com



19. References

   [BGP-MP] Bates, Chandra, Katz, and Rekhter, "Multiprotocol Extensions
   for BGP4", June 2000, RFC 2858

   [BGP-EXTCOMM] Ramachandra, Tappan, "BGP Extended Communities
   Attribute", February 2000, work in progress

   [BGP-ORF] Chen, Rekhter, "Cooperative Route Filtering Capability for
   BGP-4", March 2000, work in progress

   [BGP-RFSH] Chen, "Route Refresh Capability for BGP-4", March 2000,
   work in progress

   [BGP-RR] Bates and Chandrasekaran, "BGP Route Reflection: An
   alternative to full mesh IBGP", RFC 2796, April 2000

   [IPSEC] Kent and Atkinson, "Security Architecture for the Internet
   Protocol", November 1998, RFC 2401

   [MPLS-ARCH] Rosen, Viswanathan, and Callon, "Multiprotocol Label
   Switching Architecture", August 1999, work in progress




Rosen, et al.                                                  [Page 36]


Internet Draft       draft-rosen-rfc2547bis-02.txt             July 2000


   [MPLS-BGP] Rekhter and Rosen, "Carrying Label Information in BGP4",
   January 2000, work in progress

   [MPLS-LDP] Andersson, Doolan, Feldman, Fredette, Thomas, "LDP
   Specification", October 1999, work in progress

   [MPLS-ENCAPS] Rosen, Rekhter, Tappan, Farinacci, Fedorkow, Li, and
   Conta, "MPLS Label Stack Encoding", October 1999, work in progress

   [MPLS-RSVP] Awduche, Gan, Li, Swallow, and Srinavasan, "Extensions to
   RSVP for LSP Tunnels", March 2000, work in progress

   [PASTE] Li and Rekhter, "A Provider Architecture for Differentiated
   Services and Traffic Engineering (PASTE)", RFC 2430, October 1998.


20. Full Copyright Statement

   Copyright (C) The Internet Society (2000).  All Rights Reserved.

   This document and translations of it may be copied and furnished to
   others, and derivative works that comment on or otherwise explain it
   or assist in its implementation may be prepared, copied, published
   and distributed, in whole or in part, without restriction of any
   kind, provided that the above copyright notice and this paragraph are
   included on all such copies and derivative works.  However, this
   document itself may not be modified in any way, such as by removing
   the copyright notice or references to the Internet Society or other
   Internet organizations, except as needed for the purpose of
   developing Internet standards in which case the procedures for
   copyrights defined in the Internet Standards process must be
   followed, or as required to translate it into languages other than
   English.

   The limited permissions granted above are perpetual and will not be
   revoked by the Internet Society or its successors or assigns.

   This document and the information contained herein is provided on an
   "AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING
   TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING
   BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION
   HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF
   MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.








Rosen, et al.                                                  [Page 37]