| Network Working Group | C. Cassar |
| Internet-Draft | I. Kouvelas |
| Intended status: Experimental | D. Lewis |
| Expires: February 14, 2015 | Cisco Systems |
| August 13, 2014 |
LISP Reliable Transport
draft-kouvelas-lisp-reliable-transport-01.txt
The communication between LISP ETRs and Map-Servers is based on unreliable UDP message exchange coupled with periodic message transmission in order to maintain soft state. The drawback of periodic messaging is the constant load imposed on both the ETR and the Map-Server. New use cases for LISP have increased the amount of state that needs to be communicated with requirements that are not satisfied by the current mechanism. This document introduces the use of a reliable transport for ETR to Map-Server communication in order to eliminate the periodic messaging overhead, while providing reliability, flow-control and endpoint liveness detection.
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 February 14, 2015.
Copyright (c) 2014 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 communication channel between LISP ETRs and Map-Servers is based on unreliable UDP message exchange [RFC6833]. Where required, reliability is pursued through periodic retransmissions that maintain soft state on the peer. Map-Register messages are retransmitted every minute by an ETR and the Map-Server times out its state if the state is not refreshed for three successive periods. When registering multiple EID-Prefixes, the ETR includes multiple mapping records in the Map-Register message. Packet size limitations provide an upper bound to the number of mapping records that can be placed in each Map-Register message. When the ETR has more EID-Prefixes to register than can be packed in a single Map-Register message, the mapping records for the EID-Prefixes are split across multiple Map-Register messages.
The drawback of the periodic registration is the constant load that it introduces on both the ETR and the Map-Server. The ETR uses resources to periodically build and transmit the Map-Register messages, and to process the resulting Map-Notify messages issued by the Map-Server. The Map-Server uses resources to process the received Map-Register messages, update the corresponding registration state, and build and transmit the matching Map-Notify messages. When the number of EID-Prefixes to be registered by an ETR is small, the resulting load imposed by periodic registrations may not be significant. The ETR will only transmit a single Map-Register message each period that contains a small number of mapping records.
In some LISP deployments, a large set of EID-Prefixes must be registered by each ETR (e.g. mobility, database redistribution). Use cases with a large set of EID-Prefixes behind an ETR will result in a much higher load. An example is LISP mobility deployments where EID-Prefixes are limited to host entries. ETRs may have thousands of hosts to register resulting in hundreds of Map-Register and Map-Notify messages per registration period.
A transport is required for the ETR to Map-Server communication that provides reliability, flow-control and endpoint liveness notifications. This document describes the use of TCP or SCTP as a LISP reliable transport. The initial application for the LISP reliable transport session is the support of scalable EID prefix registration. The reliable session mechanism is defined to be extensible so that it can support additional LISP communication requirements as they arise using a single reliable transport session between an ETR and a Map-Server. The use of the reliable transport session for EID prefix registration is an alternative and does not replace the existing UDP based mechanism.
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].
A single LISP reliable transport session may carry information for multiple LISP applications. One such application is the registration of EID to RLOC mappings that operates over a session between an ETR and a Map-Server. Communication over a session is based on the exchange of messages. This document defines a base set of messages to support session establishment and management. It also defines the messages for the EID to RLOC mapping registration application.
To support protocol extensibility when new applications, or extensions to existing applications are introduced, the messages are based on a TLV format.
0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Type | Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Message ID | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Message Data ... +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Message End Marker | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Reliable transport message format
The base message format does not indicate how the peer should deal with the message in cases where the message type is not supported/understood. This is best dealt with by the application. For example, in case an error notification is returned, or an expected acknowledgement message is not received, the application might choose various courses of action; from simply logging that the feature is not supported, all the way to tearing the relationship with the peer down for the feature, or for all LISP features.
The LISP router that performs the active open initiates the connection from a locally generated source transport port number to the well-known destination transport port assigned to LISP. The LISP router that performs the passive open listens on the well-known local transport port and does not qualify the remote transport port number. In the ETR to Map-Server reliable transport session, the ETR assumes the active role and the Map-Server passively accepts connections.
A single reliable transport session can be established between a pair of LISP peers to cover all communication needs. For example, an ETR that has EID prefix registrations for multiple EID instances and EID address families might only establish a single session with the Map-Server.
When using TCP and symmetric connection establishment LISP must perform collision detection and duplicate session elimination. To accomplish that, LISP peer ID messages will be exchanged between the peers once a session is established. If duplicate sessions are detected then the one that was initiated by the router with the higher ID is kept and the other session is torn down. TBD
The error notification message is used to communicate base reliable transport session communication errors. LISP applications making use of the reliable transport session and having to communicate application specific errors must define their own messages to do so. An error notification is issued when the receiver of a message does not recognize the message type or cannot parse the message contents. The notification includes the offending message type and ID and as much of the offending message data as the notification sender wishes to.
0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Type = 16 | Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Message ID | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Error Code | Reserved | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Offending Message Type | Offending Message Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Offending Message ID | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Offending Message Data ... +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Message End Marker | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Error notification message format
An error notification cannot be the offending message in another error notification and MUST NOT trigger such a message.
EID prefix registration uses the reliable transport session between an ETR and a Map-Server to communicate the ETR local EID database EID to RLOC mappings to the Map-Server. In contrast to the UDP based periodic registration, mapping information over the reliable transport session is only sent when there is new information available for the Map-Server. The Map-Server does not maintain a timer to expire registrations communicated over the reliable transport session. Instead an explicit de-registration (a registration carrying a zero TTL) is needed to delete the state maintained by the Map-Server.
The key used to identify registration mapping records in the ETR to Map-Server communication is the EID prefix. The prefix may be specified using an LCAF encoding that includes an EID instance ID.
When the reliable transport session goes down, registration mappings learned by the Map-Server are treated as periodic UDP registrations and a timer is used to expire them after 3 minutes. During this period UDP based registrations or the re-establishment of the reliable transport session and subsequent communication of a new mapping can update the EID prefix mapping state.
This section defines the LISP reliable transport session messages used to communicate local EID database registrations between the ETR and the Map-Server.
The reliable transport Registration message is used to communicate EID to RLOC mapping registrations from the ETR to the Map-Server. The Registration message uses exactly the same format as the UDP Map-Register message but instead of the IP/UDP header, the Map-Register is placed within the value section of the reliable transport TLV. A common message format is proposed to leverage the authentication features built into the UDP Map-Register message and increase code reuse.
0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Type = 17 | Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Message ID | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Map-Register message ... +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ ... Map-Register message | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Message End Marker | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Registration message format
The Acknowledgement message is sent from the Map-Server to the ETR to confirm successful registration of an EID prefix previously communicated by a reliable transport session Registration message. The Registration Acknowledgement message does not carry a mapping record (the map servers view of the mapping). This is accomplished by the LISP reliable transport Map Notification message.
0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Type = 18 | Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Message ID | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | EID-Prefix-AFI | EID-Prefix ... +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Message End Marker | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Registration Acknowledgement message format
Negative acknowledgement sent from the Map-Server to the ETR to indicate that the registration of a specific EID prefix was rejected. The ETR must keep track of the fact that the registration of the EID prefix was rejected by the Map-Server and be prepared to re-register the mapping when requested through a failed Registration Refresh request.
0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Type = 19 | Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Message ID | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Rejection code | EID-Prefix-AFI | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | EID-Prefix ... +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Message End Marker | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Registration Rejected message format
Sent by the Map-Server to the ETR to request the re-transmission of EID prefix database mapping Registration messages.
0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Type = 20 | Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Message ID | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |R| Reserved | EID-Prefix-AFI | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | EID-Prefix ... +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Message End Marker | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Registration Refresh message format
Mapping Notification messages communicate the Map-Server view of the mapping for an EID prefix and no longer serve as a registration acknowledgement. Mapping Notifications do not need message level authentication as they are received over a reliable transport session to a known Map-Server. Note that reliable transport Mapping Notification messages do not reuse the UDP Map-Notify message format.
0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Type = 21 | Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Message ID | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Mapping Record ... +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ ... Mapping Record | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Message End Marker | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Registration message format
The ETR operates the following per EID prefix, per MS state machine that defines the reliable transport EID prefix registration behavior.
There are five states:
The following events drive the state transitions:
The state machine is:
+--------------------+--------------------------------------+ | | Prev State | | Event +-------------------+------------------+ | | No state | Periodic | +--------------------+-------------------+------------------+ | DB creation | -> Periodic | N/A | | [session down] | A1 | | +--------------------+-------------------+------------------+ | DB creation | -> AckWait | N/A | | [session up] | A2 | | +--------------------+-------------------+------------------+ | DB deletion | N/A | -> No state | | | | A3 | +--------------------+-------------------+------------------+ | DB change | N/A | - | | | | A1 | +--------------------+-------------------+------------------+ | Session up | - | -> Stable | | | | A4 | +--------------------+-------------------+------------------+ | Session down | - | N/A | | | | | +--------------------+-------------------+------------------+ | Recv Refresh | - | N/A | | | | | +--------------------+-------------------+------------------+ | Recv Refresh | - | N/A | | [rejected] | | | +--------------------+-------------------+------------------+ | Recv ACK | - | N/A | | | | | +--------------------+-------------------+------------------+ | Recv Rejection | - | N/A | | | | | +--------------------+-------------------+------------------+ | Timer | N/A | - | | | | A5 | +--------------------+-------------------+------------------+
xTR per EID prefix per MS state machine
+-----------------+-----------------------------------------------+ | | Prev State | | Event +---------------+---------------+---------------+ | | Stable | AckWait | Rejected | +-----------------+---------------+---------------+---------------+ | DB creation | N/A | N/A | N/A | | | | | | +-----------------+---------------+---------------+---------------+ | DB deletion | -> No state | -> No state | -> No state | | | A6 | A6 | | +-----------------+---------------+---------------+---------------+ | DB change | -> AckWait | - | -> AckWait | | | A2 | A2 | A2 | +-----------------+---------------+---------------+---------------+ | Session up | N/A | N/A | N/A | | | | | | +-----------------+---------------+---------------+---------------+ | Session down | -> Periodic | -> Periodic | -> Periodic | | | A7 | A7 | A7 | +-----------------+---------------+---------------+---------------+ | Recv Refresh | -> AckWait | - | -> AckWait | | | A2 | A2 | A2 | +-----------------+---------------+---------------+---------------+ | Recv Refresh | - | - | -> AckWait | | [rejected] | | A2 | A2 | +-----------------+---------------+---------------+---------------+ | Recv ACK | - | -> Stable | -> AckWait | | | | | A2 | +-----------------+---------------+---------------+---------------+ | Recv Rejection | -> Rejected | -> Rejected | - | | | | | | +-----------------+---------------+---------------+---------------+ | Timer | N/A | N/A | N/A | | | | | | +-----------------+---------------+---------------+---------------+
xTR per EID prefix per MS state machine
Action descriptions:
All timer start actions must be jittered.
When the reliable transport session is established the state machine moves into the Stable state without first registering the EID prefix over the reliable transport session. The subsequent refresh issued by the Map-Server will trigger the registration message to be sent. This model will allow future optimisations where the Map-Server may retain registration state from a previous instantiation of the reliable transport session with the ETR and only request the refresh of EID prefix state beyond some negotiated session progress marker.
Aa Map-Server authentication key change is treated as a DB change event and will result in triggering a new Registration message to be transmitted.
Received registrations create/update or delete mapping state.
A refresh for an unspecified prefix is sent when a session is first established to obtain the complete database contents from the ETR.
Refresh for rejected registrations sent (R bit set) when a new EID prefix is configured on the Map-Server.
Rejection sent to the ETR when an EID prefix that is registered is deconfigured.
Rejected Refresh (R bit set) sent when authentication for an EID prefix changes followed by a Rejection for existing registrations which fail authentication following change.
Mapping Notification message sent whenever the mapping for a registered or more specific prefix for which notifications are requested changes. ETR acknowledgement or rejection messaging for Mapping Notification is not required because the ETR decides how to process the message based on the registered mapping information. If the mapping information changes the resulting registration will trigger a new Mapping Notification message from the Map-Server.
The LISP reliable transport session SHOULD be authenticated. On controlled RLOC networks that can guarantee that the source RLOC address of data packets cannot be spoofed, the authentication check can be a source address validation on the reliable transport packets. When the RLOC network does not provide such guarantees, reliable transport authentication SHOULD be used. Implementations SHOULD support the TCP Authentication Option (TCP-AO) [RFC5925] and SCTP Authenticated Chunks [RFC4895].
Assignment of new LISP reliable transport message types is done according to the "IETF Review" model defined in [RFC5266].
The initial content of the registry should be as follows.
Type Name Reference ----------- ---------------------------------------- -------------- 0-15 Reserved This document 16 Error Notification This document 17 Registration Message This document 18 Registration Acknowledgement Message This document 19 Registration Rejected Message This document 20 Registration Refresh Message This document 21 Mapping Notification Message This document 22-30 Reserved for EID membership distribution TBD 31-64999 Unassigned 65000-65535 Reserved for Experimental Use
TCP port 4342 already reserved for LISP CONS that is now obsolete. Repurpose for reliable transport over TCP. Reserve an SCTP port.
The authors would like to thank Noel Chiappa, Dino Farinacci, Jesper Skriver, Johnson Leong, Andre Pelletier and Les Ginsberg for their contributions to this document.
| [I-D.ietf-lisp-lcaf] | Farinacci, D., Meyer, D. and J. Snijders, "LISP Canonical Address Format (LCAF)", Internet-Draft draft-ietf-lisp-lcaf-05, May 2014. |
| [RFC2119] | Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. |
| [RFC5266] | Devarapalli, V. and P. Eronen, "Secure Connectivity and Mobility Using Mobile IPv4 and IKEv2 Mobility and Multihoming (MOBIKE)", BCP 136, RFC 5266, June 2008. |
| [RFC6830] | Farinacci, D., Fuller, V., Meyer, D. and D. Lewis, "The Locator/ID Separation Protocol (LISP)", RFC 6830, January 2013. |
| [RFC6833] | Fuller, V. and D. Farinacci, "Locator/ID Separation Protocol (LISP) Map-Server Interface", RFC 6833, January 2013. |