Requirements for Header Compression over MPLS

Network Working Group                                             J. Ash
Request for Comments: 4247                                      B. Goode
Category: Informational                                          J. Hand
                                                                    AT&T
                                                                R. Zhang
                                                              BT Infonet
                                                           November 2005


             Requirements for Header Compression over MPLS

Status of This Memo

   This memo provides information for the Internet community.  It does
   not specify an Internet standard of any kind.  Distribution of this
   memo is unlimited.

Copyright Notice

   Copyright (C) The Internet Society (2005).

Abstract

   Voice over IP (VoIP) typically uses the encapsulation
   voice/RTP/UDP/IP.  When MPLS labels are added, this becomes
   voice/RTP/UDP/IP/MPLS-labels.  For an MPLS VPN, the packet header is
   typically 48 bytes, while the voice payload is often no more than 30
   bytes, for example.  Header compression can significantly reduce the
   overhead through various compression mechanisms, such as enhanced
   compressed RTP (ECRTP) and robust header compression (ROHC).  We
   consider using MPLS to route compressed packets over an MPLS Label
   Switched Path (LSP) without compression/decompression cycles at each
   router.  This approach can increase the bandwidth efficiency as well
   as processing scalability of the maximum number of simultaneous flows
   that use header compression at each router.  In this document, we
   give a problem statement, goals and requirements, and an example
   scenario.

Table of Contents

   1. Introduction ....................................................2
   2. Problem Statement ...............................................3
      2.1. Specification of Requirements ..............................4
   3. Goals and Requirements ..........................................5
   4. Candidate Solution Methods and Needs ............................6
   5. Example Scenario ................................................7
   6. Security Considerations .........................................8
   7. Normative References ............................................8
   8. Informative References ..........................................9
   9. Acknowledgements ...............................................10

1.  Introduction

   Voice over IP (VoIP) typically uses the encapsulation
   voice/RTP/UDP/IP.  When MPLS labels [MPLS-ARCH] are added, this
   becomes voice/RTP/UDP/IP/MPLS-labels.  For an MPLS Virtual Private
   Network (VPN) (e.g., [MPLS-VPN]), the packet header is at least 48
   bytes, while the voice payload is often no more than 30 bytes, for
   example.  The interest in header compression (HC) is to exploit the
   possibility of significantly reducing the overhead through various
   compression mechanisms, such as with enhanced compressed RTP [ECRTP]
   or robust header compression [ROHC], and also to increase scalability
   of HC.  We consider using MPLS to route compressed packets over an
   MPLS Label Switched Path (LSP) without compression/decompression
   cycles at each router.  Such an HC over MPLS capability can increase
   bandwidth efficiency as well as the processing scalability of the
   maximum number of simultaneous flows that use HC at each router.

   To implement HC over MPLS, the ingress router/gateway would have to
   apply the HC algorithm to the IP packet, the compressed packet routed
   on an MPLS LSP using MPLS labels, and the compressed header would be
   decompressed at the egress router/gateway where the HC session
   terminates.  Figure 1 illustrates an HC over MPLS session established
   on an LSP that crosses several routers, from R1/HC --> R2 --> R3 -->
   R4/HD, where R1/HC is the ingress router where HC is performed, and
   R4/HD is the egress router where header decompression (HD) is done.
   HC of the RTP/UDP/IP header is performed at R1/HC, and the compressed
   packets are routed using MPLS labels from R1/HC to R2, to R3, and
   finally to R4/HD, without further decompression/recompression cycles.
   The RTP/UDP/IP header is decompressed at R4/HD and can be forwarded
   to other routers, as needed.

                    _____
                   |     |
                   |R1/HC| Header Compression (HC) Performed
                   |_____|
                      |
                      | voice/compressed-header/MPLS-labels
                      V
                    _____
                   |     |
                   | R2  |
                   |_____|
                      |
                      | voice/compressed-header/MPLS-labels
                      V
                    _____
                   |     |
                   | R3  |
                   |_____|
                      |
                      | voice/compressed-header/MPLS-labels
                      V
                    _____
                   |     |
                   |R4/HD| Header Decompression (HD) Performed
                   |_____|

            Figure 1.  Example of Header Compression over MPLS
                           over Routers R1-->R4

   In the example scenario, HC therefore takes place between R1 and R4,
   and the MPLS path transports voice/compressed-header/MPLS-labels
   instead of voice/RTP/UDP/IP/MPLS-labels, typically saving 30 octets
   or more per packet.  The MPLS label stack and link-layer headers are
   not compressed.  A signaling method is needed to set up a
   correspondence between the ingress and egress routers of the HC over
   MPLS session.

   In Section 2 we give a problem statement, in Section 3 we give goals
   and requirements, and in Section 5 we give an example scenario.

2.  Problem Statement

   As described in the introduction, HC over MPLS can significantly
   reduce the header overhead through HC mechanisms.  The need for HC
   may be important on low-speed links where bandwidth is more scarce,
   but it could also be important on backbone facilities, especially
   where costs are high (e.g., some global cross-sections).  VoIP
   typically will use voice compression mechanisms (e.g., G.729) on
   low-speed and international routes, in order to conserve bandwidth.
   With HC, significantly more bandwidth could be saved.  For example,
   carrying uncompressed headers for the entire voice load of a large
   domestic network with 300 million or more calls per day could consume
   on the order of about 20-40 gigabits per second on the backbone
   network for headers alone.  This overhead could translate into
   considerable bandwidth capacity.

   The claim is often made that once fiber is in place, increasing the
   bandwidth capacity is inexpensive, nearly 'free'.  This may be true
   in some cases; however, on some international cross-sections,
   especially, facility/transport costs are very high and saving
   bandwidth on such backbone links is very worthwhile.  Decreasing the
   backbone bandwidth is needed in some areas of the world where
   bandwidth is very expensive.  It is also important in almost all
   locations to decrease the bandwidth consumption on low-speed links.
   So although bandwidth is getting cheaper, the value of compression
   does not go away.  It should be further noted that IPv6 will increase
   the size of headers, and therefore increase the importance of HC for
   RTP flows.

   Although hop-by-hop HC could be applied to decrease bandwidth
   requirements, that implies a processing requirement for compression-
   decompression cycles at every router hop, which does not scale well
   for large voice traffic loads.  The maximum number of compressed RTP
   (cRTP) flows is about 30-50 for a typical customer premise router,
   depending upon its uplink speed and processing power, while the need
   may exceed 300-500 for a high-end case.  Therefore, HC over MPLS
   seems to be a viable alternative to get the compression benefits
   without introducing costly processing demands on the intermediate
   nodes.  By using HC over MPLS, routers merely forward compressed
   packets without doing a decompression/recompression cycle, thereby
   increasing the maximum number of simultaneous compressed flows that
   routers can handle.

   Therefore, the proposal is to use existing HC techniques, together
   with MPLS labels, to make the transport of the RTP/UDP/IP headers
   more efficient over an MPLS network.  However, at this time, there
   are no standards for HC over MPLS, and vendors have not implemented
   such techniques.

2.1.  Specification of Requirements

   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 [KEY].

3.  Goals and Requirements

   The goals of HC over MPLS are as follows:

   a. provide more efficient voice transport over MPLS networks,
   b. increase the scalability of HC to a large number of flows,
   c. not significantly increase packet delay, delay variation, or loss
      probability, and
   d. leverage existing work through use of standard protocols as much
      as possible.

   Therefore the requirements for HC over MPLS are as follows:

   a. MUST use existing protocols (e.g., [ECRTP], [ROHC]) to compress
      RTP/UDP/IP headers, in order to provide for efficient transport,
      tolerance to packet loss, and resistance to loss of session
      context.
   b. MUST allow HC over an MPLS LSP, and thereby avoid hop-by-hop
      compression/decompression cycles (e.g., [HC-MPLS-PROTO]).
   c. MUST minimize incremental performance degradation due to increased
      delay, packet loss, and jitter.
   d. MUST use standard protocols to signal context identification and
      control information (e.g., [RSVP], [RSVP-TE], [LDP]).
   e. Packet reordering MUST NOT cause incorrectly decompressed packets
      to be forwarded from the decompressor.

   It is necessary that the HC method be able to handle out-of-sequence
   packets.  MPLS [MPLS-ARCH] enables 4-byte labels to be appended to IP
   packets to allow switching from the ingress Label Switching Router
   (LSR) to the egress LSP on an LSP through an MPLS network.  However,
   MPLS does not guarantee that packets will arrive in order at the
   egress LSR, since a number of things could cause packets to be
   delivered out of sequence.  For example, a link failure could cause
   the LSP routing to change, due perhaps to an MPLS fast reroute taking
   place, or to the Interior Gateway Protocol (IGP) and Label
   Distribution Protocol (LDP) converging to another route, among other
   possible reasons.  Other causes could include IGP reroutes due to
   'loose hops' in the LSP, or BGP route changes reflecting back into
   IGP reroutes.  HC algorithms may be able to handle reordering
   magnitudes on the order of about 10 packets, which may make the time
   required for IGP reconvergence (typically on the order of seconds)
   untenable for the HC algorithm.  On the other hand, MPLS fast reroute
   may be fast enough (on the order of 50 ms or less) for the HC
   algorithm to handle packet reordering.  The issue of reordering needs
   to be further considered in the development of the HC over MPLS
   solution.

   Resynchronization and performance also needs to be considered, since
   HC over MPLS can sometimes have multiple routers in the LSP.
   Tunneling an HC session over an MPLS LSP with multiple routers in the
   path will increase the round-trip delay and the chance of packet
   loss, and HC contexts may be invalidated due to packet loss.  The HC
   error recovery mechanism can compound the problem when long round-
   trip delays are involved.

4.  Candidate Solution Methods and Needs

   [cRTP] performs best with very low packet error rates on all hops of
   the path.  When the cRTP decompressor context state gets out of synch
   with the compressor, it will drop packets associated with the context
   until the two states are resynchronized.  To resynchronize context
   state at the two ends, the decompressor transmits the CONTEXT_STATE
   packet to the compressor, and the compressor transmits a FULL_HEADER
   packet to the decompressor.

   [ECRTP] uses mechanisms that make cRTP more tolerant to packet loss,
   and ECRTP thereby helps to minimize the use of feedback-based error
   recovery (CONTEXT_STATE packets).  ECRTP is therefore a candidate
   method to make HC over MPLS more tolerant of packet loss and to guard
   against frequent resynchronizations.  ECRTP may need some
   implementation adaptations to address the reordering requirement in
   Section 3 (requirement e), since a default implementation will
   probably not meet the requirement.  ECRTP protocol extensions may be
   required to identify FULL_HEADER, CONTEXT_STATE, and compressed
   packet types.  [cRTP-ENCAP] specifies a separate link-layer packet
   type defined for HC.  Using a separate link-layer packet type avoids
   the need to add extra bits to the compression header to identify the
   packet type.  However, this approach does not extend well to MPLS
   encapsulation conventions [MPLS-ENCAP], in which a separate link-
   layer packet type translates into a separate LSP for each packet
   type.  In order to extend ECRTP to HC over MPLS, each packet type
   defined in [ECRTP] would need to be identified in an appended packet
   type field in the ECRTP header.

   [ROHC] is also very tolerant of packet loss, and therefore is a
   candidate method to guard against frequent resynchronizations.  ROHC
   also achieves a somewhat better level of compression as compared to
   ECRTP.  ROHC may need some implementation adaptations to address the
   reordering requirement in Section 3 (requirement e), since a default
   implementation will probably not meet the requirement (see
   [ROHC-REORD]).  ROHC already has the capability to identify the
   packet type in the compression header, so no further extension is
   needed to identify packet type.

   Extensions to MPLS signaling may be needed to identify the LSP from
   HC to HD egress point, negotiate the HC algorithm used and protocol
   parameters, and negotiate the Session Context IDs (SCIDs) space
   between the ingress and egress routers on the MPLS LSP.  For example,
   new objects may need to be defined for [RSVP-TE] to signal the SCID
   spaces between the ingress and egress routers, and the HC algorithm
   used to determine the context; these HC packets then contain the SCID
   identified by using the RSVP-TE objects.  It is also desirable to
   signal HC over MPLS tunnels with the Label Distribution Protocol
   [LDP], since many RFC 2547 VPN [MPLS-VPN] implementations use LDP as
   the underlying LSP signaling mechanism, and LDP is very scalable.
   However, extensions to LDP may be needed to signal SCIDs between
   ingress and egress routers on HC over MPLS LSPs.  For example,
   'targeted LDP sessions' might be established for signaling SCIDs, or
   perhaps methods described in [LDP-PWE3] to signal pseudo-wires and
   multipoint-to-point LSPs might be extended to support signaling of
   SCIDs for HC over MPLS LSPs.  The specific MPLS signaling protocol
   extensions to support these approved requirements need to be
   developed as a well-coordinated separate document in the appropriate
   IETF working groups.  The IETF needs to support a coordinated process
   for the two solution documents, though they are in separate areas.

5.  Example Scenario

   As illustrated in Figure 2, many VoIP flows are originated from
   customer sites, which are served by routers R1, R2, and R3, and
   terminated at several large customer call centers, which are served
   by R5, R6, and R7.  R4 is a service-provider router, and all VoIP
   flows traverse R4.  It is essential that the R4-R5, R4-R6, and R4-R7
   low-speed links all use HC to allow a maximum number of simultaneous
   VoIP flows.  To allow processing at R4 to handle the volume of
   simultaneous VoIP flows, it is desired to use HC over MPLS for these
   flows.  With HC over MPLS, R4 does not need to do HC/HD for the flows
   to the call centers, enabling more scalability of the number of
   simultaneous VoIP flows with HC at R4.

        voice/C-HDR/MPLS-labels ______ voice/C-HDR/MPLS-labels
   R1/HC---------------------->|      |-----------------------> R5/HD
                               |      |
        voice/C-HDR/MPLS-labels|      |voice/C-HDR/MPLS-labels
   R2/HC---------------------->|  R4  |-----------------------> R6/HD
                               |      |
        voice/C-HDR/MPLS-labels|      |voice/C-HDR/MPLS-labels
   R3/HC---------------------->|______|-----------------------> R7/HD

                    Note: HC    = header compression
                          C-HDR = compressed header
                          HD    = header decompression

        Figure 2.  Example Scenario for Application of HC over MPLS

6.  Security Considerations

   The high processing load of HC makes HC a target for denial-of-
   service attacks.  For example, an attacker could send a high-
   bandwidth data stream through a network, with the headers in the data
   stream marked appropriately to cause HC to be applied.  This would
   use large amounts of processing resources on the routers performing
   compression and decompression, and these processing resources might
   then be unavailable for other important functions on the router.
   This threat is not a new threat for HC, but is addressed and
   mitigated by HC over MPLS.  That is, by reducing the need for
   performing compression and decompression cycles, as proposed in this
   document, the risk of this type of denial-of-service attack is
   reduced.

7.  Normative References

   [cRTP]          Casner, S. and V. Jacobson, "Compressing IP/UDP/RTP
                   Headers for Low-Speed Serial Links", RFC 2508,
                   February 1999.

   [cRTP-ENCAP]    Engan, M., Casner, S., Bormann, C., and T. Koren, "IP
                   Header Compression over PPP", RFC 3544, July 2003.

   [ECRTP]         Koren, T., Casner, S., Geevarghese, J., Thompson, B.,
                   and P. Ruddy, "Enhanced Compressed RTP (CRTP) for
                   Links with High Delay, Packet Loss and Reordering",
                   RFC 3545, July 2003.

   [KEY]           Bradner, S., "Key words for use in RFCs to Indicate
                   Requirement Levels", RFC 2119, March 1997.

   [LDP]           Andersson, L., Doolan, P., Feldman, N., Fredette, A.,
                   and B. Thomas, "LDP Specification", RFC 3036, January
                   2001.

   [MPLS-ARCH]     Rosen, E., Viswanathan, A., and R. Callon,
                   "Multiprotocol Label Switching Architecture", RFC
                   3031, January 2001.

   [ROHC]          Bormann, C., et al., "RObust Header Compression
                   (ROHC): Framework and four profiles: RTP, UDP, ESP,
                   and uncompressed ", RFC 3095, July 2001.

   [RSVP]          Braden, R., Zhang, L., Berson, S., Herzog, S., and S.
                   Jamin, "Resource ReSerVation Protocol (RSVP) --
                   Version 1 Functional Specification", RFC 2205,
                   September 1997.

   [RSVP-TE]       Awduche, D., Berger, L., Gan, D., Li, T., Srinivasan,
                   V., and G. Swallow, "RSVP-TE: Extensions to RSVP for
                   LSP Tunnels", RFC 3209, December 2001.

8.  Informative References

   [HC-MPLS-PROTO] Ash, G., Goode, B., Hand, J., Jonsson, L-E., Malis,
                   A., and R. Zhang, "Protocol Extensions for Header
                   Compression over MPLS", work in progress.

   [LDP-PWE3]      Martini, L., El-Aawar, N., Smith, T., and G. Heron,
                   "Pseudowire Setup and Maintenance using the Label
                   Distribution Protocol", work in progress.

   [MPLS-ENCAP]    Rosen, E., Tappan, D., Fedorkow, G., Rekhter, Y.,
                   Farinacci, D., Li, T., and A. Conta, "MPLS Label
                   Stack Encoding", RFC 3032, January 2001.

   [MPLS-VPN]      Rosen, E. and Y. Rekhter, "BGP/MPLS VPNs", RFC 2547,
                   March 1999.

   [ROHC-REORD]    Pelletier, G., Jonsson, L-E., and K. Sandlund,
                   "RObust Header Compression (ROHC): ROHC over Channels
                   that can Reorder Packets", work in progress.

9.  Acknowledgements

   The authors wish to thank the following people (in alphabetical
   order) for their helpful comments and suggestions:  Loa Andersson,
   Scott Brim, Thomas Eriksson, Victoria Fineberg, Lars-Erik Jonsson,
   Allison Mankin, Colin Perkins, Kristofer Sandlund, and Magnus
   Westerlund.

Authors' Addresses

   Jerry Ash
   AT&T
   Room MT D5-2A01
   200 Laurel Avenue
   Middletown, NJ 07748, USA
   Phone: +1 732-420-4578
   EMail: gash@att.com


   Bur Goode
   AT&T
   Phone: + 1 203-341-8705
   EMail: bgoode@att.com


   Jim Hand
   AT&T
   Room MT A2-1A03
   200 Laurel Avenue
   Middletown, NJ 07748, USA
   Phone: +1 732-420-3017
   EMail: jameshand@att.com


   Raymond Zhang
   BT Infonet
   2160 E. Grand Ave.
   El Segundo, CA 90025 USA
   EMail: raymond.zhang@bt.infonet.com

Full Copyright Statement

   Copyright (C) The Internet Society (2005).

   This document is subject to the rights, licenses and restrictions
   contained in BCP 78, and except as set forth therein, the authors
   retain all their rights.

   This document and the information contained herein are provided on an
   "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS
   OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET
   ENGINEERING TASK FORCE DISCLAIM 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.

Intellectual Property

   The IETF takes no position regarding the validity or scope of any
   Intellectual Property Rights or other rights that might be claimed to
   pertain to the implementation or use of the technology described in
   this document or the extent to which any license under such rights
   might or might not be available; nor does it represent that it has
   made any independent effort to identify any such rights.  Information
   on the procedures with respect to rights in RFC documents can be
   found in BCP 78 and BCP 79.

   Copies of IPR disclosures made to the IETF Secretariat and any
   assurances of licenses to be made available, or the result of an
   attempt made to obtain a general license or permission for the use of
   such proprietary rights by implementers or users of this
   specification can be obtained from the IETF on-line IPR repository at
   http://www.ietf.org/ipr.

   The IETF invites any interested party to bring to its attention any
   copyrights, patents or patent applications, or other proprietary
   rights that may cover technology that may be required to implement
   this standard.  Please address the information to the IETF at ietf-
   ipr@ietf.org.

Acknowledgement

   Funding for the RFC Editor function is currently provided by the
   Internet Society.