in Index   Prev   Next

RFC 4031

Service Requirements for Layer 3 Provider Provisioned Virtual Private Networks (PPVPNs)

Pages: 50
Part 2 of 3 – Pages 15 to 36
First   Prev   Next

Top   ToC   RFC4031 - Page 15   prevText

5. Customer Requirements

This section captures additional requirements from a customer perspective.

5.1. VPN Membership (Intranet/Extranet)

When an extranet is formed, a customer agent from each of the organizations first approves addition of a site to an extranet VPN as a business decision between the parties involved. The solution SHOULD provide a means for these organizations to control extranet communication involving the L3VPN exchange of traffic and routing information.
Top   ToC   RFC4031 - Page 16

5.2. Service Provider Independence

Customers MAY require VPN service that spans multiple administrative domains or service provider networks. Therefore, a VPN service MUST be able to span multiple AS and SP networks, but still act and appear as a single, homogeneous VPN from a customer point of view. A customer might also start with a VPN provided in a single AS with a certain SLA but then ask for an expansion of the service, spanning multiple ASes/SPs. In this case, as well as for all kinds of multi- AS/SP VPNs, VPN service SHOULD be able to deliver the same SLA to all sites in a VPN regardless of the AS/SP to which it homes.

5.3. Addressing

A customer requires support from an L3VPN for the following addressing IP assignment schemes: o Customer-assigned, non-unique, or [RFC1918] private addresses o Globally unique addresses obtained by the customer o Globally unique addresses statically assigned by the L3VPN service provider o On-demand, dynamically assigned IP addresses (e.g., DHCP), irrespective of whether the access is temporary (e.g., remote) or permanent (e.g., dedicated) In the case of combined L3VPN service with non-unique or private addresses and Internet access, mechanisms that permit the exchange of traffic between the customer's address space and the global unique Internet address space MAY be supported. For example, NAT is employed by many customers and by some service providers today to meet this need. A preferred solution would be to assign unique addresses, either IPv4 or IPv6; however, some customers do not want to renumber their networks.

5.4. Routing Protocol Support

There SHOULD be no restriction on the routing protocols used between CE and PE routers, or between CE routers. At least the following protocols MUST be supported: static routing, IGP protocols such as RIP, OSPF, IS-IS, and BGP [L3VPN-FR].

5.5. Quality of Service and Traffic Parameters

QoS is expected to be an important aspect of an L3VPN service for some customers. QoS requirements cover scenarios involving an intranet, an extranet, and shared access between a VPN site and the Internet.
Top   ToC   RFC4031 - Page 17

5.5.1. Application-Level QoS Objectives

A customer is concerned primarily that the L3VPN service provides his or her applications with the QoS and level of traffic so that the applications perform acceptably. Voice, interactive video, and multimedia applications are expected to require the most stringent QoS. These real-time applications are sensitive to delay, delay variation, loss, availability, and/or reliability. Another set of applications, including some multimedia and interactive video applications, high-performance web browsing, and file transfer intensive applications, requires near real time performance. Finally, best effort applications are not sensitive to degradation, that is they are elastic and can adapt to conditions of degraded performance. The selection of appropriate QoS and service type to meet specific application requirements is particularly important to deal with periods of congestion in an SP network. Sensitive applications will likely select per-flow Integrated service (Intserv) with precise SLA guarantees measured on a per-flow basis. On the other hand, non- sensitive applications will likely rely on a Diffserv class-based QoS. The fundamental customer application requirement is that an L3VPN solution MUST support both the Intserv QoS model for selected individual flows and Diffserv for aggregated flows. A customer application SHOULD experience consistent QoS independent of the access network technology used at different sites connected to the same VPN.

5.5.2. DSCP Transparency

The Diffserv Code Point (DSCP) set by a user as received by the ingress CE SHOULD be capable of being relayed transparently to the egress CE (see section 2.6.2 of [RFC3270] and [Y.1311.1]). Although RFC 2475 states that interior or boundary nodes within a DS domain can change the DSCP, customer VPNs MAY have other requirements, such as o applications that use the DSCP in a manner differently from the DSCP solution supported by the SP network(s), o customers using more DSCPs within their sites than the SP network(s) supports, o support for a carrier's carrier service in which one SP is the customer of another L3VPN SP. Such an SP should be able to resell VPN service to his or her VPN customers independently of the DSCP mapping solution supported by the carrier's carrier SP.
Top   ToC   RFC4031 - Page 18
   Note that support for DSCP transparency has no implication on the QoS
   or SLA requirements.  If an SP supports DSCP transparency, then that
   SP needs to carry only the DSCP values across its domain but MAY map
   the received DSCP to some other value for QoS support across its

5.6. Service-Level Specification/Agreement

Most customers simply want their applications to perform well. An SLA is a vehicle for customer recourse in the event that SP(s) do not perform or manage a VPN service well in a measurable sense. Therefore, when purchasing service under an SLA, a customer agent MUST have access to the measures from the SP(s) that support the SLA.

5.7. Customer Management of a VPN

A customer MUST have a means to view the topology, operational state, order status, and other parameters associated with his or her VPN. Most aspects of management information about CE devices and customer attributes of an L3VPN manageable by an SP SHOULD be capable of being configured and maintained by an authenticated, authorized customer agent. However, some aspects, such as encryption keys, SHALL NOT be readable nor writable by management systems. A customer agent SHOULD be able to make dynamic requests for changes to traffic parameters. A customer SHOULD be able to receive real- time response from the SP network in response to these requests. One example of such service is a "Dynamic Bandwidth management" capability that enables real-time response to customer requests for changes of allocated bandwidth allocated to his or her VPN [Y.1311.1]. A customer who may not be able to afford the resources to manage his own sites SHOULD be able to outsource the management of the entire VPN to the SP(s) supporting the VPN network.

5.8. Isolation

These features include traffic and routing information exchange isolation, similar to that obtained in VPNs based on Layer 1 and Layer 2 (e.g., private lines, FR, or ATM) [MPLSSEC].
Top   ToC   RFC4031 - Page 19

5.9. Security

The suite of L3VPN solutions SHOULD support a range of security related features. Higher levels of security services, such as edge- to-edge encryption, authentication, or replay attack, should be supported. More details on customer requirements for security are described in [VPNSEC]. Security in an L3VPN service SHOULD be as transparent as possible to the customer, with the obvious exception of support for remote or temporary user access, as detailed in section 5.11.2. L3VPN customers MUST be able to deploy their own internal security mechanisms in addition to those deployed by the SP, in order to secure specific applications or traffic at a granularity finer than that on a site-to-site basis. If a customer requires QoS support in an L3VPN, then this request MUST be communicated to the SP either by using unencrypted fields or via an agreed security association. For example, applications could send RSVP messages in support of Intserv either in the clear or encrypted with a key negotiated with the SP. Another case is that where applications using an IPsec tunnel could copy the DSCP from the encrypted IP header to the header of the tunnel's IP header.

5.10. Migration Impact

Often, customers are migrating from an already deployed private network toward one or more L3VPN solutions. A typical private network scenario is CE routers connected via real or virtual circuits. Ideally, minimal incremental cost SHOULD result during the migration period. Furthermore, if necessary, any disruption of service SHOULD also be minimized. A range of scenarios of customer migration MUST be supported. Full migration of all sites MUST be supported. Support for cases of partial migration is highly desirable [Y.1311.1] - that is, legacy private network sites that belong to the L3VPN service SHOULD still have L3 reachability to the sites that migrate to the L3VPN service.

5.11. Network Access

Every L3 packet exchanged between the customer and the SP over the access connection MUST appear as it would on a private network providing an equivalent service to that offered by the L3VPN.
Top   ToC   RFC4031 - Page 20

5.11.1. Physical/Link Layer Technology

L3VPNs SHOULD support a broad range of physical and link-layer access technologies, such as PSTN, ISDN, xDSL, cable modem, leased line, Ethernet, Ethernet VLAN, ATM, Frame Relay, Wireless local loop, and mobile radio access. The capacity and QoS achievable may be dependent on the specific access technology in use.

5.11.2. Temporary Access

The VPN service offering SHOULD allow both permanent and temporary access to one or more L3VPNs for authenticated users across a broad range of access technologies. Support for remote or temporary VPN access SHOULD include ISDN, PSTN dial-in, xDSL, or access via another SP network. The customer SHOULD be able to choose from alternatives for authentication of temporary access users. Choices for access authentication are SP-provided, third-party, or customer-provided authentication. A significant number of VPN users may not be permanently attached to one VPN site: in order to limit access to a VPN to authorized users, it is first necessary to authenticate them. Authentication SHALL apply as configured by the customer agent and/or SP where a specific user may be part of one or more VPNs. The authentication function SHOULD be used to invoke all actions necessary to join a user to the VPN automatically. A user SHOULD be able to access an L3VPN via a network having generic Internet access. Mobile users may move within an L3VPN site. Mobile users may also have temporary connections to different L3VPN sites within the same VPN. Authentication SHOULD be provided in both of these cases.

5.11.3. Sharing of the Access Network

In a PE-based L3VPN, if the site shares the access network with other traffic (e.g., access to the Internet), then data security in the access network is the responsibility of the L3VPN customer.

5.11.4. Access Connectivity

Various types of physical connectivity scenarios MUST be supported, such as multi-homed sites, backdoor links between customer sites, and devices homed to two or more SP networks. L3VPN solutions SHOULD support at least the types of physical or link-layer connectivity arrangements shown in Figure 2.1. Support for other physical connectivity scenarios with arbitrary topology is desirable.
Top   ToC   RFC4031 - Page 21
   Access arrangements with multiple physical or logical paths from a CE
   to other CEs and PEs MUST support redundancy and SHOULD support load
   balancing.  Resiliency uses redundancy to provide connectivity
   between a CE site and other CE sites and, optionally, other services.
   Load balancing provides a means to perform traffic engineering so
   that capacity on redundant links is used to achieve improved
   performance during periods when the redundant component(s) are

   For multi-homing to a single SP, load balancing capability SHOULD be
   supported by the PE across the CE to PE links.  For example, in case
   (a), load balancing SHOULD be provided by the two PEs over the two
   links connecting to the single CE.  In case (c), load balancing
   SHOULD be provided by the two PEs over the two links connecting to
   the two CEs.

   In addition, the load-balancing parameters (e.g., the ratio of
   traffic on the multiple load-balanced links, or the preferred link)
   SHOULD be provisionable based on customer's requirements.  The load-
   balancing capability may also be used to achieve resiliency in the
   event of access connectivity failures.  For example, in case (b) a CE
   may connect to two different SPs via diverse access networks.
   Resiliency MAY be further enhanced as shown in case (d), where CEs
   connected via a "back door" connection connect to different SPs.
   Furthermore, arbitrary combinations of the above methods, with a few
   examples shown in cases (e) and (f), should be supportable by any
   L3VPN approach.

   For multi-homing to multiple SPs, load balancing capability MAY also
   be supported by the PEs in the different SPs (clearly, this is a more
   complex type of load balancing to realize, requiring policy and
   service agreements between the SPs to interoperate).
Top   ToC   RFC4031 - Page 22
                   +----------------                    +---------------
                   |                                    |
                +------+                            +------+
      +---------|  PE  |                  +---------|  PE  |
      |         |router|                  |         |router| SP network
      |         +------+                  |         +------+
   +------+         |                  +------+         |
   |  CE  |         |                  |  CE  |         +---------------
   |device|         |   SP network     |device|         +---------------
   +------+         |                  +------+         |
      |         +------+                  |         +------+
      |         |  PE  |                  |         |  PE  |
      +---------|router|                  +---------|router| SP network
                +------+                            +------+
                    |                                   |
                    +----------------                   +---------------
                   (a)                                 (b)
                    +----------------                  +---------------
                    |                                  |
   +------+     +------+               +------+     +------+
   |  CE  |-----|  PE  |               |  CE  |-----|  PE  |
   |device|     |router|               |device|     |router| SP network
   +------+     +------+               +------+     +------+
      |             |                     |             |
      | Backdoor    |                     | Backdoor    +---------------
      | link        |   SP network        | link        +---------------
      |             |                     |             |
   +------+     +------+               +------+     +------+
   |  CE  |     |  PE  |               |  CE  |     |  PE  |
   |device|-----|router|               |device|-----|router| SP network
   +------+     +------+               +------+     +------+
                    |                                   |
                    +----------------                   +---------------
                   (c)                                  (d)
Top   ToC   RFC4031 - Page 23
                   +----------------                    +---------------
                   |                                    |
  +------+     +------+                +------+     +------+
  |  CE  |-----|  PE  |                |  CE  |-----|  PE  |
  |device|     |router|                |device|     |router| SP network
  +------+\\   +------+                +------+\\   +------+
     |     \\       |                     |     \\       |
     |Back  \\      |                     |Back  \\
     |door   \\     |   SP network        |door   \\
     |link    \\    |                     |link    \\    |
  +------+     +------+               +------+     +------+
  |  CE  |     |  PE  |               |  CE  |     |  PE  |
  |device|-----|router|               |device|-----|router| SP network
  +------+     +------+               +------+     +------+
                   |                                   |
                   +----------------                   +---------------
                  (e)                                 (f)

         Figure 2.1.  Representative types of access arrangements

5.12. Service Access

Customers MAY also require access to other services, as described in this section.

5.12.1. Internet Access

Customers SHOULD be able to have L3VPN and Internet access across the same access network for one or more of the customer's sites. Customers SHOULD be able to direct Internet traffic from the set of sites in the L3VPN to one or more customer sites that have firewalls, other security-oriented devices, and/or NATs that process all traffic between the Internet and the customer's VPN. L3 VPN Customers SHOULD be able to receive traffic from the Internet addressed to a publicly accessible resource that is not part of the VPN, such as an enterprise's public web server. As stated in section 5.3, if a customer L3VPN employs private or non-unique IP addresses, then network address translation (NAT) or a similar mechanism MUST be provided either by the customer or the SP in order to allow traffic exchange with devices outside the customer's L3VPN.
Top   ToC   RFC4031 - Page 24

5.12.2. Hosting, Application Service Provider

A customer SHOULD be able to access hosting, other application services, or other Application Service Providers (ASP) over an L3 L3VPN service. This MAY require that an ASP participate in one or more VPNs with the customers that use such a service.

5.12.3. Other Services

In conjunction with a VPN service, a customer MAY also wish to have access to other services, such as DNS, FTP, HTTP, NNTP, SMTP, LDAP, VoIP, NAT, LDAP, Videoconferencing, Application sharing, E-business, Streaming, E-commerce, Directory, Firewall, etc. The resources that implement these services could be physically dedicated to each VPN. If the resources are logically shared, then they MUST have access separated and isolated between VPNs in a manner consistent with the L3VPN solution to meet this requirement.

5.13. Hybrid VPN Service Scenarios

Intranet or extranet customers have a number of reasons for wanting hybrid networks that involve more than one VPN solution type. These include migration, mergers, extranet customers with different VPN types, the need for different capabilities between different sets of sites, temporary access, and different availability of VPN solutions as provided by different service providers. The framework and solution approaches SHOULD include provisions for interworking, interconnection, and/or reachability between different L3VPN solutions in a way that does not overly complicate provisioning, management, scalability, or performance.

6. Service Provider Network Requirements

This section describes requirements from a service provider perspective.

6.1. Scalability

[RFC3809] lists projections of L3VPN sizing and scalability requirements and metrics related to specific solutions.
Top   ToC   RFC4031 - Page 25

6.2. Addressing

As described in section 4.2, SPs MUST have support for public and private IP addresses, IPv4 and IPv6, for both unicast and multicast. In order to support this range of addressing schemes, SPs require the following support from L3VPN solutions. An L3VPN solution MUST be able to assign blocks of addresses from its own public IP address space to L3VPN customer sites so that advertisement of routes to other SPs and other sites aggregates efficiently. An L3VPN solution MUST be able to use address assignments made by a customer. These customer-assigned addresses may be public or private. If private IP addresses are used, an L3VPN solution MUST provide a means for an SP to translate such addresses to public IP addresses for communication with other VPNs by using overlapping addresses or the Internet.

6.3. Identifiers

A number of identifiers MAY be necessary for SP use in management, control, and routing protocols. Requirements for at least the following identifiers are known. An SP domain MUST be uniquely identified at least within the set of all interconnected SP networks when supporting a VPN that spans multiple SPs. Ideally, this identifier should be globally unique (e.g., an AS number). An identifier for each VPN SHOULD be unique, at least within each SP's network. Ideally, the VPN identifier SHOULD be globally unique to support the case where a VPN spans multiple SPs (e.g., [RFC2685]). A CE device SHOULD have a unique identifier, at least within each SP's network. A PE device SHOULD have a unique identifier, at least within each SP's network. The identifier of a device interconnecting SP networks MUST be unique within the set of aforementioned networks. Each site interface SHOULD have a unique identifier, at least within each PE router supporting such an interface.
Top   ToC   RFC4031 - Page 26
   Each tunnel SHOULD have a unique identifier, at least within each
   router supporting the tunnel.

6.4. Discovering VPN Related Information

Configuration of CE and PE devices is a significant task for a service provider. Solutions SHOULD strive to contain methods that dynamically allow VPN information to be discovered (or learned) by the PE and/or CE to reduce configuration complexity. The following specific requirements apply to intra- and inter-provider VPNs [VPNDISC]. Every device involved in a VPN SHALL be able to identify and authenticate itself to other devices in the VPN. After learning the VPN membership, the devices SHOULD be able to exchange configuration information securely. The VPN information MUST include at least the IP address of the PE and may be extensible to provide additional information. Each device in a VPN SHOULD be able to determine which other devices belong to the same VPN. Such a membership discovery scheme MUST prevent unauthorized access and allow authentication of the source. Distribution of VPN information SHOULD be limited to those devices involved in that VPN. In the case of a PE-based VPN, a solution SHOULD support the means for attached CEs to authenticate each other and verify that the SP's VPN network is correctly configured. The mechanism SHOULD respond to VPN membership changes in a timely manner. This is no longer than the provisioning timeframe, typically on the order of minutes, and could be as short as the timeframe required for "rerouting", typically on the order of seconds. Dynamically creating, changing, and managing multiple VPN assignments to sites and/or customers is another aspect of membership that MUST be addressed in an L3VPN solution.

6.5. SLA and SLS Support

Typically, a Service Provider offering an L3VPN service commits to specific Service Level Specifications (SLS) as part of a contract with the customer, as described in section 4.4 and [RFC3809]. Such a Service Level Agreement (SLA) implies SP requirements for measuring Specific Service Level Specifications (SLS) for quality, availability, response time, and configuration intervals.
Top   ToC   RFC4031 - Page 27

6.6. Quality of Service (QoS) and Traffic Engineering

A significant aspect of an L3VPN is support for QoS. Since an SP has control over the provisioning of resources and configuration of parameters in at least the PE and P devices and, in some cases, in the CE device as well, the onus is on the SP to provide either managed QoS access service, or edge-to-edge QoS service, as defined in section 4.3.2. Each L3VPN approach MUST describe the traffic engineering techniques available for an SP to meet the QoS objectives. These descriptions of traffic engineering techniques SHOULD quantify scalability and achievable efficiency. Traffic engineering support MAY be on an aggregate or per-VPN basis. QoS policies MUST not be impacted by security mechanisms. For example, Diffserv policies MUST not be impacted by the use of IPSec tunnels using the mechanisms explained in RFC 2983 [RFC2983]. As stated in RFC 2475, a mapping function from customer provided Diffserv marking to marking used in an SP network should be provided for L3 VPN services. If a customer requires DSCP transparency, as described in section 5.5.2, an L3VPN service MUST deliver the same value of DSCP field in the IP header received from the customer to the egress demarcation of the destination.

6.7. Routing

The distribution of reachability and routing policy SHOULD be constrained to the sites that are members of the VPN. Optionally, the exchange of such information MAY use some form of authentication (e.g., MD5). Functions to isolate the SP network and customer VPNs from anomalous routing behavior from a specific set of customer sites SHOULD be provided. Examples of such functions are controls for route flap dampening, filters that accept only prefixes configured for a specific CE, a maximum number of routes accepted for each CE, or a maximum rate at which route updates can be received from a CE. When VPN customers use overlapping non-unique IP addresses, the solution MUST define a means to distinguish between such overlapping addresses on a per-VPN basis.
Top   ToC   RFC4031 - Page 28
   Furthermore, the solution SHOULD provide an option that either allows
   or prevents advertisement of VPN routes to the Internet.

   Ideally, the choice of an SP's IGP SHOULD not depend on the routing
   protocol(s) used between PE and CE routers in a PE-based VPN.

   Furthermore, it is desirable that an SP SHOULD have a choice
   regarding the IGP routing protocol.

   The additional routing burden that an SP must carry should be
   articulated in each specific L3VPN solution.

6.8. Isolation of Traffic and Routing

The internal structure of an L3VPN network SHOULD not be visible to outside networks (e.g., the Internet or any connected VPN). From a high-level SP perspective, a PE-based L3VPN MUST isolate the exchange of traffic and routing information to only those sites that are authenticated and authorized members of a VPN. In a CE-based VPN, the tunnels that connect the sites effectively meet this isolation requirement if both traffic and routing information flow over the tunnels. An L3VPN solution SHOULD provide a means to meet L3VPN QoS SLA requirements that isolates VPN traffic from the effects of traffic offered by non-VPN customers. Also, L3VPN solutions SHOULD provide a means to isolate the effects that traffic congestion produced by sites as part of one VPN can have on another VPN.

6.9. Security

This section contains requirements related to securing customer flows; providing authentication services for temporary, remote, or mobile users; and protecting service provider resources involved in supporting an L3VPN. More detailed security requirements are provided in [VPNSEC].

6.9.1. Support for Securing Customer Flows

In order to meet the general requirement for providing a range of security options to a customer, each L3VPN solution MUST clearly spell out the configuration options that can work together and how they can do so.
Top   ToC   RFC4031 - Page 29
   When a VPN solution operates over a part of the Internet, it should
   support a configurable option to support one or more of the following
   standard IPsec methods for securing a flow for a specified subset of
   a customer's VPN traffic:

   o  Confidentiality, so that only authorized devices can decrypt it
   o  Integrity, to ensure that the data has not been altered
   o  Authentication, to ensure that the sender is indeed who he or she
      claims to be
   o  Replay attack prevention.

   The above functions SHOULD be applicable to "data traffic" of the
   customer, which includes the traffic exchanged between sites between
   temporary users and sites, and even between temporary users.  It
   SHOULD also be possible to apply these functions to "control
   traffic", such as routing protocol exchanges, that are not
   necessarily perceived by the customer but are nevertheless essential
   to maintain his or her VPN.

   Furthermore, such security methods MUST be configurable between
   different end points, such as CE-CE, PE-PE, and CE-PE.  It is also
   desirable to configure security on a per-route or per-VPN basis

   A VPN solution MAY support one or more encryption schemes, including
   AES, and 3DES.  Encryption, decryption, and key management SHOULD be
   included in profiles as part of the security management system.

6.9.2. Authentication Services

A service provider MUST provide authentication services in support of temporary user access requirements, as described in section 5.11.2. Furthermore, traffic exchanged within the scope of VPN MAY involve several categories of equipment that must cooperate to provide the service [Y.1311.1]. These network elements can be CE, PE, firewalls, backbone routers, servers, management stations, etc. These network elements learn about each other's identity, either via manual configuration or via discovery protocols, as described in section 6.4. When network elements must cooperate, these network elements SHALL authenticate peers before providing the requested service. This authentication function MAY also be used to control access to network resources. The peer identification and authentication function described above applies only to network elements participating in the VPN. Examples include:
Top   ToC   RFC4031 - Page 30
   -  traffic between a CE and a PE,
   -  traffic between CEs belonging to the same VPN,
   -  CE or PE routers dealing with route announcements for a VPN,
   -  policy decision point [RFC3198] and a network element, and
   -  management station and an SNMP agent.

   For a peer authentication function, each L3VPN solution SHOULD
   describe where necessary, how it shall be implemented, how secure it
   must be, and the way to deploy and maintain identification and
   authentication information necessary to operate the service.

6.9.3. Resource Protection

Recall from the definitions in section 3.3 that a site can be part of an intranet with sites from the only same organization, can be part of an extranet involving sites from other organizations, can have access to the Internet, or can have any combination of these scopes of communication. Within these contexts, a site might be subject to various attacks coming from different sources. Potential sources of attack include: - users connected to the supporting public IP backbone, - users from the Internet, and - users from temporary sites belonging to the intranet and/or extranet VPN the site is part of. Security threats and risks that a site may encounter include the following: - Denial of service, for example mail spamming, access connection congestion, TCP SYN attacks, and ping attacks - Intrusion attempts, which may eventually lead to denial of service (e.g., a Trojan horse attack). Additional threat scenarios are defined in [VPNSEC]. An L3VPN solution MUST state how it addresses each potential threat scenario. The devices in the L3VPN network must provide some means of reporting intrusion attempts to the service provider resources.

6.10. Inter-AS (SP)VPNs

The scenario for VPNs spanning multiple Autonomous Systems (AS) or Service Providers (SP) requires standard solutions. The scenario where multiple ASes are involved is the most general case and is therefore the one described here. The scenarios of concern are the CE-based and PE-based L3VPNs defined in section 3.
Top   ToC   RFC4031 - Page 31
   In each scenario, all applicable SP requirements, such as traffic and
   routing isolation, SLAs, management, security, and provisioning.
   MUST be preserved across adjacent ASes.  The solutions MUST describe
   the inter-SP network interface, encapsulation method(s), routing
   protocol(s), and all applicable parameters [VPNIW].

   An essential pre-condition for an inter-AS VPN is an agreement
   between the ASes involved that spells out at least trust, economic,
   and management responsibilities.

   The overall scalability of the VPN service MUST allow the L3VPN
   service to be offered across potentially hundreds of SPs, with the
   overall scaling parameters per SP given in [RFC3809].

6.10.1. Routing Protocols

If the link between ASes is not trusted, routing protocols running between those ASes MUST support some form of authentication. For example, the TCP option for carrying an MD5 digest may be used to enhance security for BGP [RFC2385]. BGP MUST be supported as the standard inter-AS routing protocol to control the path taken by L3VPN traffic.

6.10.2. Management

The general requirements for managing a single AS apply to a concatenation of ASes. A minimum subset of such capabilities as follows: - Diagnostic tools (e.g., ping, traceroute) - Secured access to one AS management system by another - Configuration request and status query tools - Fault notification and trouble-tracking tools

6.10.3. Bandwidth and QoS Brokering

When a VPN spans multiple ASes, a brokering mechanism is desired that requests certain SLA parameters, such as bandwidth and QoS, from the other domains and/or networks involved in transferring traffic to various sites. Although bandwidth and QoS brokering across multiple ASes is not common in today's networks, these may be desirable for maintaining SLAs in inter-AS VPNs. This section describes requirements for features that would facilitate these mechanisms. The objective is that a solution SHOULD be able to determine whether a set of ASes can establish and guarantee uniform QoS in support of an L3VPN.
Top   ToC   RFC4031 - Page 32
   The brokering mechanism can be a manual one, for example, in which
   one provider requests from another a specific set of bandwidth and
   QoS parameters for traffic going to and from a specific set of sites.
   The mechanism could also be an automated one where a device
   dynamically requests and receives certain bandwidth and SLA/QoS
   parameters.  For instance, in the case of an L3VPN over MPLS, a PE
   may negotiate the label for different traffic classes to reach a PE
   residing in a neighboring AS.  Or, it might be a combination of both.
   For additional detailed requirements on the automated approach, see

   Brokering on a per VPN basis is not desirable as this approach would
   not scale.  A solution MUST provide some means to aggregate QoS and
   bandwidth brokering requests between ASes.  One method could be for
   SPs to make an agreement specifying the maximum amount of bandwidth
   for specific QoS parameters for all VPN customers using the SP
   network.  Alternatively, such aggregation might be on a per
   hierarchical tunnel basis between PE routers in different ASes
   supporting an L3VPN service [TE-INTERAS].

6.10.4. Security Considerations

If a tunnel traverses multiple SP networks and passes through an unsecured SP, POP, NAP, or IX, then security mechanisms MUST be employed. These security mechanisms include encryption, authentication, and resource protection, as described in section 6.9, and security management, as covered in section 7.5. For example, a provider should consider using both authentication and encryption for a tunnel used as part of an L3VPN that traverses another service provider's network.

6.11. L3VPN Wholesale

The architecture MUST support the possibility of one service provider offering VPN service to another service provider. Another example is when one service provider sells L3VPN service at wholesale to another service provider, who then resells that VPN service to his or her customers. The wholesaler's VPN MUST be transparent to the addressing and routing used by the reseller. Support for additional levels of hierarchy (for example, three levels at which a reseller can again resell the VPN service to yet another VPN provider) SHOULD be provided. The Carrier's Carrier scenario is the term used in this document for this category of L3VPN wholesale (although some scenarios of Inter-
Top   ToC   RFC4031 - Page 33
   AS/Inter-Provider VPN could possibly fall in this L3VPN wholesale
   category, too).  Various carrier's carrier scenarios should be
   supported, such as when

   -  the customer carriers do not operate L3VPN services for their
   -  the customer carriers operate L3VPN services for their clients,
      but these services are not linked with the L3VPN service offered
      by the Carrier's Carrier and
   -  the customer carriers operate L3VPN services for their clients,
      and these services are linked with the L3VPN service offered by
      the Carrier's Carrier ("Hierarchical VPNs" scenario).

6.12. Tunneling Requirements

Connectivity between CE sites or PE devices in the backbone SHOULD use a range of tunneling technologies, such as L2TP, IPSEC, GRE, IP- in-IP, and MPLS. To set up tunnels between routers, every router MUST support static configuration for tunneling and MAY support a tunnel setup protocol. If employed, a tunnel establishment protocol SHOULD be capable of conveying information such as the following: - Relevant identifiers - QoS/SLA parameters - Restoration parameters - Multiplexing identifiers - Security parameters There MUST be a means to monitor the following aspects of tunnels: - Statistics, such as amount of time spent in the up and down state. - Count of transitions between the up and down state. - Events, such as transitions between the up and down states. The tunneling technology used by the VPN Service Provider and its associated mechanisms for tunnel establishment, multiplexing, and maintenance MUST meet the requirements on scaling, isolation, security, QoS, manageability, etc.

6.13. Support for Access and Backbone Technologies

This section describes requirements for aspects of access and backbone network technologies from an SP point of view.
Top   ToC   RFC4031 - Page 34
   Some SPs MAY desire that a single network infrastructure suffices for
   all services, public IP, VPNs, traffic engineering, and
   differentiated services [L2VPN].

6.13.1. Dedicated Access Networks

Ideally, the L3VPN service SHOULD be independent of physical, link layer, or even network technology of the access network. However, the characteristics of access networks MUST be accounted for when the QoS aspects of SLAs for VPN service offerings are specified.

6.13.2. On-Demand Access Networks

Service providers SHOULD be able to support temporary user access, as described in section 5.11.2, by using dedicated or dial-in access network technology. L3VPN solutions MUST support the case where a VPN user directly accesses the VPN service through an access network connected to the service provider. They MUST also describe how they can support the case where one or more other service provider networks are used for access to the service provider supporting the L3VPN service. Ideally, all information necessary to identify and authenticate users for an intranet SHOULD be stored and maintained by the customer. In an extranet, one customer SHOULD be able to maintain the authentication server, or the customers involved in the extranet MAY choose to outsource the function to a service provider. Identification and authentication information could be made available to the service provider for controlling access, or the service provider may query a customer maintained server. Furthermore, one SP may act as access for the SP providing the VPN service. If the access SP performs identification and authentication on behalf of the VPN SP, an agreement MUST be reached on a common specification. Support for at least the following authentication protocols SHALL be supported: PAP, CHAP, and EAP, as they are currently used in a wide range of equipment and services.
Top   ToC   RFC4031 - Page 35

6.13.3. Backbone Networks

Ideally, the backbone interconnecting SP, PE, and P devices SHOULD be independent of physical and link layer technology. Nevertheless, the characteristics of backbone technology MUST be taken into account when specifying the QoS aspects of SLAs for VPN service offerings.

6.14. Protection, Restoration

When primary and secondary access connections are available, an L3VPN solution MUST provide restoration of access connectivity whenever the primary access link from a CE site to a PE fails. This capability SHOULD be as automatic as possible, that is, the traffic should be directed over the secondary link soon after failure of the primary access link is detected. Furthermore, reversion to the primary link SHOULD be dynamic, if configured to do so [VPN-NEEDS]. As mentioned in section 5.11.4, in the case of multi-homing, the load balancing capability MAY be used to achieve a degree of redundancy in the network. In the case of failure of one or more (but not all) of the multi-homed links, the load balancing parameters MAY be dynamically adjusted to redirect the traffic rapidly from the failed link(s) to the surviving links. Once the failed link(s) is (are) restored, the original provisioned load balancing ratio SHOULD be restored to its value prior to the failure. An SP SHOULD be able to deploy protection and restoration mechanisms within his or her backbone infrastructure to increase reliability and fault tolerance of the VPN service offering. These techniques SHOULD be scalable, and therefore should strive not to perform such function in the backbone on a per-VPN basis. Appropriate measurements and alarms that indicate how well network protection and restoration mechanisms are performing MUST be supported.

6.15. Interoperability

Service providers are interested in interoperability in at least the following scenarios: - Facilitating use of PE and managed CE devices within a single SP network. - Implementing L3VPN services across two or more interconnected SP networks.
Top   ToC   RFC4031 - Page 36
   -  Achieving interworking or interconnection between customer sites
      using different L3VPN approaches or different implementations of
      the same approach.

   Each approach MUST describe whether any of the above objectives can
   be met.  If an objective can be met, the approach MUST describe how
   such interoperability could be achieved.  In particular, the approach
   MUST describe the inter-solution network interface, encapsulation
   method(s), routing protocol(s), security, isolation, management, and
   all other applicable aspects of the overall VPN solution provided

6.16. Migration Support

Service providers MUST have a graceful means to migrate a customer with minimal service disruption on a site-by-site basis to an L3VPN approach. If L3VPN approaches can interwork or interconnect, then service providers MUST have a graceful means to migrate a customer with minimal service disruption on a site-by-site basis whenever interworking or interconnection is changed.

(page 36 continued on part 3)

Next Section