| Softwire Working Group | I. Farrer |
| Internet-Draft | Q. Sun |
| Intended status: Standards Track | Deutsche Telekom AG |
| Expires: January 7, 2016 | July 6, 2015 |
Multiple BR Tunnel Endpoint Addresses
draft-farrer-softwire-br-multiendpoints-01
As 'softwire' based approaches for providing IPv4 services to IPv6 networks
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in [RFC2119].
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at http://datatracker.ietf.org/drafts/current/.
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 material or to cite them other than as "work in progress."
This Internet-Draft will expire on January 7, 2016.
Copyright (c) 2015 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
The Softwire WG has developed a number of IPv4-over-IPv6 transition mechanisms based on a hub-and-spoke network architecture, with the Border Router (BR) functioning as the hub. Although the schema for configuring a BR varies according to the mechanism being implemented (using either a per-subscriber rule or an algorithmic mapping rule), vendor's implementations differ in their approach to the provisioning of tunnel endpoints on the BR. In some implementations, a single address must be used for terminating all clients traffic, some implementations require every client to use a different tunnel endpoint and some employ a 'hybrid' approach, allowing for the clients to have unique BR addresses, and other clients to be arbrarily grouped with multiple clients using the same BR tunnel endpoint address.
From an operator's standpoint, this difference creates a provisioning problem: The values of the parameters with which softwire initiators need to be provisioned vary depending on the BR's tunnel endpoint provisioning approach. In a multi-vendor environment, this becomes unmanagable and significantly complicates migration of users and geo-redundancy between BR's from different vendors.
This document proposes that the 'hybrid' approach for BR tunnel endpoints offers the most flexibility and should be the model implemented in softwire concentrators.
In this implementation approach, only a single IPv6 address is used as the BR's IPv6 tunnel address. This address is provisioned to all softwire initiators to use as the destination for their IPv4-in-IPv6 tunneled traffic, and is used by the BR as the source address for encapsulated traffic being sent to Softwire initiators.
The obvious advantage is that all clients can be provisioned with the same address for the BR. In smaller scale deployments, this may be sufficient for an operator's needs. In larger scale deployments, this schema makes it difficult to design and plan for the grouping of users and geographical failover.
Only supporting a single IPv6 BR tunnel endpoint address available has another limitation: it is not possbile to differentiate client's tunneled traffic based on BR's address in the encapsulating IPv6 packet.
In this implementation approach, each client is provisioned with a unique /128 address to use as the destination address for their tunneled traffic. The BR is configured with exactly as many unique tunnel endpoint addresses as there are softwire initiators.
The main disadvantage with this approach is that it places an additional provisioning overhead on the operator to ensure that each client has a unique BR address. When provisioning lw4o6 using DHCPv6 as described in [I-D.ietf-softwire-map-dhcp], maintaining per-client DHCPv6 entries is a necessary provisioning overhead. But when dynamic provisionig of lw4o6 clients is used [RFC7341], per-client entries in the provisioning system are no longer required, as the IPv4 addresses for clients are taken from an address pool. However, if each client still needs to be provisioned with a unique tunnel endpoint address associated with the IPv4 address it has been allocated, then the provisioning model becomes complex and potentially unworkable.
This document proposes that a BR SHOULD support per-client IPv6 tunnel endpoints, but not mandate that these addresses are unique. E.g., the endpoints can be all the same, or unique, or any combination of unique and shared.
By implementing multiple IPv6 tunnel endpoint addresses in this 'mixed' mode, the BR can support different classes of users, grouped through their tunnel endpoint address. Grouping clients based on a common tunnel endpoint address makes it simple for intermediate IPv6 network elements to identify client's traffic group by examining the encapsulating IPv6 header, e.g. so that traffic forwarding policies can be applied.
It also allows for flexible, anycast based geographical resilience models where each BR supports a primary group of users and a secondary group, differentiated by the tunnel endpoint address. Traffic is flexibly routed through auto-routing protocols and Equal-Cost Multipath (ECMP).
This document describes a method that enables one Border Router to serve in such a mix mode. The BR's mapping/binding table must hold an additional "BR source IPv6 address" field for each Softwire Initiator it is configured to support. The IPv6 addresses of that field can be all the same or not, based on the operational requirements.
This mechanism can be applied to lw4over6 [I-D.ietf-softwire-lw4over6], MAP-E [I-D.ietf-softwire-map] and MAP-T [I-D.ietf-softwire-map-t].
DISCUSSION - Is this necessary for MAP BRs, or can this already be supported?
Existing BRs implementing lw4o6, MAP-E or MAP-T are provisioned with a set of rules defining packet processing behaviour. The rule/binding table on the Border Router only contains the mapping between the IPv6 and IPv4 address and source ports for the Softwire Initiators. In this mechanism, the rule/binding table is extended to include the IPv6 tunnel address(es) configured on the BR as another field. The Softwire Initiators' IPv6-IPv4 mapping rules are then linked to the related BR's IPv6 tunnel addresses. As such, it is possible for one BR to serve multiple groups of Softwire Initiators, independently from each other.
On receiving an IPv6 packet, the BR MUST use both the source and the destination IPv6 addresses as input parameters to search for a matching entry in the mapping rule table, instead of only using the source IPv6 address/prefix information. If a successful match is made, the encapsulated/translated IPv4 packet is then verified as documented in related Softwire mechanisms.
When the BR receives a packet from the IPv4 Internet, it looks up for the matching entry using the destination IPv4 address and port number. The BR MUST also retrieve the associated BR's IPv6 address to use for the encapsulating packet's source IPv6 address. Depending on the implemented mechanism, the mapping rule for constructing the destination IPv6 address will need to be retrieved as normal.
The rest of encapsulation/decapsulation/translation process is inline with the related mechanisms.
The Softwire Initiator's behavior is identical to that in the related mechanisms. That is, when it receives an IPv4-in-IPv6 packet it checks the source and destination addresses against its configuration. The source address of the packet MUST match the BR's tunnel endpoint address configured on the client.
Border Routers need to be provisioned with multiple sets of tunnel endpoint IPv6 addresses, IPv4-IPv6 mapping rules for Softwire Initiators and routable IPv4 prefixes. The provisioning mechanisms could include NETCONF, TR-69 or out-of-band static configuration. This mechanism is out of scope for this document.
BRs implementing this mechanism can be deployed using IPv6 anycast to achieve high availability. Since multiple IPv6 anycast addresses can be configured on the BR as tunnel endpoint addresses, a BR is able to serve one primary domain while serving other domains as backup. The BR advertises the IPv6 anycast prefix(es), as well as the routable IPv4 prefix(es). ECMP can be used to leverage for stateless load-balancing across multiple BRs.
However, as the reachable IPv4 customer prefixes are being advertised by all instances serving that domain simultanously, IPv4 traffic which ingresses the network will, by default, use the cluster which has the lowest routing metric to the ingress point in the network. This may results in different paths for egress and ingress traffic. Whilst stateless and per-subscriber state softwire mechansims (described in [RFC6269]) don't require the ingress/egress traffic paths to be symmetrical, it may be desirable for an operator to engineer this way for effective capacity planning. The exact mechanism for achieving this will be dependant on the network's topology and how the operator is utilizes equal-cost multipath based load balancing.
NOTE: Another possible consideration is that as there is an additional lookup action that needs to be carried out for packets at the BR, there may be a packet processing overhead.
TBD
This document does not include an IANA request.
The authors would like to thank Madhusuhdan Vadde for contributions to this work.
| [RFC2119] | Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. |
| [RFC2473] | Conta, A. and S. Deering, "Generic Packet Tunneling in IPv6 Specification", RFC 2473, December 1998. |