Internet Engineering Task Force                      Bruce Thompson
   Audio/Video Transport Working Group                     Tmima Koren
   INTERNET-DRAFT                                             Dan Wing
   EXPIRES: February 2005                                Cisco Systems
                                                       September, 2004


               Tunneling Multiplexed Compressed RTP ("TCRTP")
                      draft-ietf-avt-tcrtp-08.txt


Status of this memo

   This document is an Internet-Draft and is in full conformance with
   all provisions of Section 5 of RFC3667.

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

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

   By submitting this Internet-Draft, I certify that any applicable
   patent or other IPR claims of which I am aware have been disclosed,
   and any of which I become aware will be disclosed, in accordance
   with RFC 3668.

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

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

Copyright Notice

   Copyright (C) The Internet Society (2004).  All Rights Reserved.

Abstract

   This document describes a method to improve the bandwidth
   utilization of RTP streams over network paths that carry multiple
   Real-time Transport Protocol (RTP) streams in parallel between two
   endpoints, as in voice trunking. The method combines standard
   protocols that provide compression, multiplexing, and tunneling over
   a network path to reduce the bandwidth used when multiple RTP
   streams are carried over that path.


INTERNET-DRAFT    Tunneled Multiplexed Compressed RTP    August, 2004


Table of Contents

1.   Introduction....................................................3
1.1.  Is Bandwidth Costly?..........................................3
1.2.  Overview of Protocols.........................................3
1.3.  Document Focus................................................3
1.4.  Choice of Enhanced CRTP.......................................4
1.5.  Reducing TCRTP Overhead.......................................4
2.   Protocol Operation and Recommended Extensions...................4
2.1.  Header Compression: ECRTP.....................................5
2.1.1.  Synchronizing ECRTP States...................................5
2.1.2.  Out-of-Order Packets.........................................6
2.2.  Multiplexing: PPP Multiplexing................................6
2.2.1.  PPP Multiplex Transmitter Modifications for Tunneling........6
2.2.2.  Tunneling Inefficiencies.....................................8
2.3.  Tunneling: L2TP...............................................8
2.3.1.  Tunneling and DiffServ.......................................8
2.4.  Encapsulation Formats.........................................8
3.   Bandwidth Efficiency............................................9
3.1.  Multiplexing gains............................................9
3.2.  Packet loss rate.............................................10
3.3.  Bandwidth calculation for Voice and Video Applications.......10
3.3.1.  Voice Bandwidth Calculation Example.........................12
3.3.2.  Voice Bandwidth Comparison Table............................12
3.3.3.  Video Bandwidth Calculation Example.........................13
3.3.4.  TCRTP over ATM..............................................13
3.3.5.  TCRTP over non-ATM networks.................................14
4.   Example implementation of TCRTP................................14
4.1.  Suggested PPP and L2TP negotiation for TCRTP.................16
4.2.  PPP negotiation TCRTP........................................16
4.2.1.  LCP negotiation.............................................16
4.2.2.  IPCP negotiation............................................16
4.3.  L2TP negotiation.............................................17
4.3.1.  Tunnel Establishment........................................17
4.3.2.  Session Establishment.......................................17
4.3.3.  Tunnel Tear Down............................................18
5.   IANA Considerations............................................18
6.   Security Considerations........................................18
7.   Acknowledgements...............................................19
8.   References.....................................................19
9.   Authors' Addresses.............................................20
10.  Copyright Notice...............................................21
11.  Disclaimers....................................................21

Koren, Thompson, Wing                                         [Page 2]


INTERNET-DRAFT    Tunneled Multiplexed Compressed RTP    August, 2004



1.   Introduction

  This document describes a way to combine existing protocols for
  compression, multiplexing, and tunneling to save bandwidth for some
  RTP applications.

1.1.  Is Bandwidth Costly?

  On certain links, such as customer access links, the cost of
  bandwidth is widely acknowledged to be a significant concern.
  Protocols such as CRTP (Compressed RTP, [CRTP]) are well suited to
  help bandwidth inefficiencies of protocols such as VoIP over these
  links.

   Unacknowledged by many, however, is the cost of long-distance WAN
   links.  While some voice-over-packet technologies such as Voice over
   ATM (VoAAL2, [I.363.2]) and Voice over MPLS provide bandwidth
   efficiencies because both technologies lack IP, UDP, and RTP
   headers, neither VoATM nor VoMPLS provide direct access to voice-
   over-packet services available to Voice over IP.  Thus, goals of WAN
   link cost reduction are met at the expense of lost interconnection
   opportunities to other networks.

   TCRTP solves the VoIP bandwidth discrepancy, especially for large
   voice trunking applications.

1.2.  Overview of Protocols

   Header compression is accomplished using Enhanced CRTP (ECRTP,
   [ECRTP]). ECRTP is an enhancement to classical CRTP [CRTP] that
   works better over long delay links, such as the end-to-end tunneling
   links described in this document.  This header compression reduces
   the IP, UDP, and RTP headers.

   Multiplexing is accomplished using PPP Multiplexing [PPP-MUX].

   Tunneling PPP is accomplished by using L2TP [L2TPv3].

   CRTP operates link-by-link; that is, to achieve compression over
   multiple router hops, CRTP must be employed twice on each router --
   once on ingress, again on egress.  In contrast, TCRTP described in
   this document does not require any additional per-router processing
   to achieve header compression -- instead, headers are compressed
   end-to-end, saving bandwidth on all intermediate links.

1.3.  Document Focus


Koren, Thompson, Wing                                         [Page 3]


INTERNET-DRAFT    Tunneled Multiplexed Compressed RTP    August, 2004


   This document is primarily concerned with bandwidth savings for
   Voice over IP (VoIP) applications over high-delay networks.
   However, the combinations of protocols described in this document
   can be used to provide similar bandwidth savings for other RTP
   applications such as video, and bandwidth savings are included for a
   sample video application.

1.4.  Choice of Enhanced CRTP

  CRTP [CRTP] describes the use of RTP header compression on an
  unspecified link layer transport, but typically PPP is used.  For
  CRTP to compress headers, it must be implemented on each PPP link.  A
  lot of context is required to successfully run CRTP, and memory and
  processing requirements are high, especially if multiple hops must
  implement CRTP to save bandwidth on each of the hops.  At higher line
  rates, CRTP's processor consumption becomes prohibitively expensive.

   To avoid the per-hop expense of CRTP, a simplistic solution is to
   use CRTP with L2TP to achieve end-to-end CRTP.  However, as
   described in [ECRTP], CRTP is only suitable for links with low delay
   and low loss.  However, once multiple router hops are involved,
   CRTP's expectation of low delay and low loss can no longer be met.
   Further, packets can arrive out of order.

   Therefore, this document describes the use of Enhanced CRTP [ECRTP],
   which supports high delay, both packet loss, and misordering between
   the compressor and decompressor.

1.5.  Reducing TCRTP Overhead

   If only one stream is tunneled (L2TP) and compressed (ECRTP) there
   is little bandwidth savings.  Multiplexing is helpful to amortize
   the overhead of the tunnel header over many RTP payloads.  The
   multiplexing format that is proposed by this document is PPP
   multiplexing [PPP-MUX].  See section 2.3 for details.

2.   Protocol Operation and Recommended Extensions

   This section describes how to combine three protocols: Enhanced
   CRTP, PPP Multiplexing, and L2TP Tunneling, to save bandwidth for
   RTP applications such as Voice over IP.

2.1.  Models

   TCRTP can typically be implemented in two ways.  The most
   straightforward is to implement TCRTP in the gateways terminating
   the RTP streams:

       [voice gateway]---[voice gateway]

Koren, Thompson, Wing                                         [Page 4]


INTERNET-DRAFT    Tunneled Multiplexed Compressed RTP    August, 2004


                       ^
                       |
                 TCRTP over IP

   Another way TCRTP can be implemented is with an external
   concentration device.  This device could be placed at strategic
   places in the network and could dynamically create and destroy TCRTP
   sessions without the participation of RTP-generating endpoints.

       [voice GW]\                                   /[voice GW]
       [voice GW]---[concentrator]---[concentrator]---[voice GW]
       [voice GW]/                                   \[voice GW]
                  ^                ^                ^
                  |                |                |
             RTP over IP     TCRTP over IP     RTP over IP

   Such a design also allows classical CRTP [CRTP] to be used on links
   with only a few active flows per link (where TCRTP isn't efficient;
   see section 3):

       [voice GW]\                                   /[voice GW]
       [voice GW]---[concentrator]---[concentrator]---[voice GW]
       [voice GW]/                                   \[voice GW]
                  ^                ^                ^
                  |                |                |
           CRTP over IP     TCRTP over IP     RTP over IP


2.2.  Header Compression: ECRTP

   As described in [ECRTP], classical CRTP [CRTP] is not suitable over
   long-delay WAN links commonly used when tunneling as proposed by
   this document.  Thus, ECRTP should be used instead of CRTP.

2.2.1.    Synchronizing ECRTP States

   When the compressor receives an RTP packet which has an unpredicted
   change in the RTP header, the compressor should send a
   COMPRESSED_UDP packet (described in [ECRTP]) to synchronize the
   ECRTP decompressor state.  The COMPRESSED_UDP packet updates the RTP
   context in the decompressor.

   To ensure delivery of updates of context variables, COMPRESSED_UDP
   packets should be delivered using the robust operation described in
   [ECRTP].

   As the "twice" algorithm described in [ECRTP] relies on UDP
   checksums, the IP stack on the RTP transmitter should transmit UDP

Koren, Thompson, Wing                                         [Page 5]


INTERNET-DRAFT    Tunneled Multiplexed Compressed RTP    August, 2004


   checksums. If UDP checksums are not used, the ECRTP compressor
   should use the CRTP Headers checksum described in [ECRTP].

2.2.2.    Out-of-Order Packets

   Tunneled transport does not guarantee in order delivery of packets.
   Therefore, the ECRTP decompressor must operate correctly in the
   presence of out of order packets.

   The order of packets for RTP is determined by the RTP sequence
   number.  ECRTP sends short deltas from the RTP seqno, and sends a
   full value every N packets, where N is an engineered constant tuned
   to the kind of pipe ECRTP is used for.

   By contrast, [ROHC] compresses out the sequence number and another
   layer is necessary for [ROHC] to handle out-of-order delivery of
   packets over a tunnel [REORDER].

2.3.  Multiplexing: PPP Multiplexing

   Both CRTP and ECRTP require a layer two protocol which allows
   identifying different protocols.  [PPP] is suited for this.

   When CRTP is used inside of a tunnel, the header compression
   associated with CRTP will reduce the size of the IP, UDP, and IP
   headers of the IP packet carried in the tunnel. However, the tunnel
   itself has overhead due to its IP header and the tunnel header (the
   information necessary to identify the tunneled payload). One way to
   reduce the overhead of the IP header and tunnel header is to
   multiplex multiple RTP payloads in a single tunneled packet.

   [PPP-MUX] describes an encapsulation that combines multiple PPP
   payloads into one multiplexed payload.  PPP multiplexing allows any
   supported PPP payload type to be multiplexed.  This multiplexed
   frame is then carried as a single PPPMUX payload in the IP tunnel.
   This allows multiple RTP payloads to be carried in a single IP
   tunnel packet and allows the overhead of the uncompressed IP and
   tunnel headers to be amortized over multiple RTP payloads.

   During PPP establishment of the TCRTP tunnel, only LCP and IPCP (for
   header compression) are required -- IP addresses do not need to be
   negotiated, nor is authentication necessary.  See section 4.1 for
   details.

2.3.1.    PPP Multiplex Transmitter Modifications for Tunneling

   Section 1.2 of [PPP-MUX] describes an example transmitter procedure
   that can be used to implement a PPP Multiplex transmitter. The
   transmission procedure described in this section includes a

Koren, Thompson, Wing                                         [Page 6]


INTERNET-DRAFT    Tunneled Multiplexed Compressed RTP    August, 2004


   parameter MAX-SF-LEN that is used to limit the maximum size of a PPP
   Multiplex frame.

   There are two reasons for limiting the size of a PPP Multiplex
   frame.  First, a PPPMUX frame should never exceed the MRU of a
   physical link.  Second, when a PPP session and its associated flow
   control are bound to a physical link, the MAX-SF-LEN parameter forms
   an upper limit on the amount of time a multiplex packet can be held
   before being transmitted.  When flow control for the PPP Multiplex
   transmitter is bound to a physical link, the clock rate of the
   physical link can be used to pull frames from the PPP Multiplex
   transmitter.

   This type of flow control limits the maximum amount of time a PPP
   multiplex frame can be held before being transmitted to MAX-SF-LEN /
   Link Speed.

   Tunnel interfaces are typically not bound to physical interfaces.
   Because of this, a tunnel interface has no well-known transmission
   rate associated with it. This means that flow control in the PPPMUX
   transmitter cannot rely on the clock of a physical link to pull
   frames from the multiplex transmitter. Instead, a timer must be used
   to limit the amount of time a PPPMUX frame can be held before being
   transmitted.  The timer along with the MAX-SF-LEN parameter should
   be used to limit the amount of time a PPPMUX frame is held before
   being transmitted.

   The following extensions to the PPPMUX transmitter logic should be
   made for use with tunnels. The flow control logic of the PPP
   transmitter should be modified to collect incoming payloads until
   one of two events has occurred:

          (1)  a specific number of octets, MAX-SF-LEN, has arrived at
          the multiplexer, or;

          (2)  a timer, called T, has expired.

   When either condition is satisfied, the multiplexed PPP payload is
   transmitted.

   The purpose of MAX-SF-LEN is to ensure that a PPPMUX payload does
   not exceed the MTU size of any of the possible physical links that
   the tunnel can be associated with. The value of MAX-SF-LEN should be
   less than or equal to the minimum of MRU-2(maximum size of length
   field) and 16,383 (14 bits) for all possible physical interfaces
   that the tunnel may be associated with.

   The timer T provides an upper delay bound for tunnel interfaces.
   Timer T is reset whenever a multiplexed payload is sent to the next

Koren, Thompson, Wing                                         [Page 7]


INTERNET-DRAFT    Tunneled Multiplexed Compressed RTP    August, 2004


   encapsulation layer.  The behavior of this timer is similar to
   AAL2's Timer_CU described in [I.363.2].  Each PPPMUX transmitter
   should have its own Timer T.

   The optimal values for  T will vary depending upon the rate at which
   payloads are expected to arrive at the multiplexer and the delay
   budget for the multiplexing function.  For voice applications, the
   value of T would typically be 5-10 milliseconds.

2.3.2.    Tunneling Inefficiencies

   To get reasonable bandwidth efficiency using multiplexing within an
   L2TP tunnel, multiple RTP streams should be active between the
   source and destination of an L2TP tunnel.

   If the source and destination of the L2TP tunnel are the same as the
   source and destination of the ECRTP sessions, then the source and
   destination must have multiple active RTP streams to get any benefit
   from multiplexing.

   Because of this limitation, TCRTP is mostly useful for applications
   where many RTP sessions run between a pair of RTP endpoints.  The
   number of simultaneous RTP sessions required to reduce the header
   overhead to the desired level depends on the size of the L2TP
   header.  A smaller L2TP header will result in fewer simultaneous RTP
   sessions being required to produce bandwidth efficiencies similar to
   CRTP.

2.4.  Tunneling: L2TP

   L2TP tunnels should be used to tunnel the ECRTP payloads end to end.
   L2TP includes methods for tunneling messages used in PPP session
   establishment such as NCP.  This allows [IPCP-HC] to negotiate ECRTP
   compression/decompression parameters.

2.4.1.    Tunneling and DiffServ

   RTP streams may be marked with Expedited Forwarding (EF) bits, as
   described in [EF-PHB].  When such a packet is tunneled, the tunnel
   header must also be marked for the same EF bits, as required by [EF-
   PHB].  It is important to not mix EF and non-EF traffic in the same
   EF-marked multiplexed tunnel.

2.5.  Encapsulation Formats

   The packet format for an RTP packet compressed with RTP header
   compression as defined in ECRTP is:

        +---------+---------+-------------+-----------------------+

Koren, Thompson, Wing                                         [Page 8]


INTERNET-DRAFT    Tunneled Multiplexed Compressed RTP    August, 2004


        |         |   MSTI  |             |                       |
        | Context |         |     UDP     |                       |
        |   ID    |   Link  |   Checksum  |       RTP Data        |
        |         | Sequence|             |                       |
        |  (1-2)  |   (1)   |     (0-2)   |                       |
        +---------+---------+-------------+-----------------------+


   The packet format of a multiplexed PPP packet as defined by [PPP-
   MUX] is:

        +-------+---+------+-------+-----+   +---+------+-------+-----+
        | Mux   |P L|      |       |     |   |P L|      |       |     |
        | PPP   |F X|Len1  |  PPP  |     |   |F X|LenN  |  PPP  |     |
        | Prot. |F T|      | Prot. |Info1| ~ |F T|      | Prot. |InfoN|
        | Field |          | Field1|     |   |          |FieldN |     |
        | (1)   |1-2 octets| (0-2) |     |   |1-2 octets| (0-2) |     |
        +-------+----------+-------+-----+   +----------+-------+-----+


   The combined format used for TCRTP with a single payload is all of
   the above packets concatenated.  Here is an example with one
   payload:

        +------+-------+----------+-------+-------+-----+-------+----+
        | IP   | Mux   |P L|      |       |       | MSTI|       |    |
        |header| PPP   |F X|Len1  |  PPP  |Context|     | UDP   |RTP |
        | (20) | Proto |F T|      | Proto |  ID   | Link| Cksum |Data|
        |      | Field |          | Field1|       | Seq |       |    |
        |      | (1)   |1-2 octets| (0-2) | (1-2) | (1) | (0-2) |    |
        +------+-------+----------+-------+-------+-----+-------+----+
               |<------------- IP payload ------------------------->|
                       |<----- PPPmux payload --------------------->|

   If the tunnel contains multiplexed traffic, multiple "PPPMux
   payload"s are transmitted in one IP packet.

3.   Bandwidth Efficiency

   The expected bandwidth efficiency attainable with TCRTP depends upon
   a number of factors.  These factors include multiplexing gain,
   expected packet loss rate across the network, and rates of change of
   specific fields within the IP and RTP headers.  This section also
   describes how TCRTP significantly enhances bandwidth efficiency for
   voice over IP over ATM.

3.1.  Multiplexing gains


Koren, Thompson, Wing                                         [Page 9]


INTERNET-DRAFT    Tunneled Multiplexed Compressed RTP    August, 2004


   Multiplexing reduces the overhead associated with the layer 2 and
   tunnel headers.  Increasing the number of CRTP payloads combined
   into one multiplexed PPP payload increases multiplexing gain.  As
   traffic increases within a tunnel, more payloads are combined in one
   multiplexed payload.  This will increase multiplexing gain.

3.2.  Packet loss rate

   Loss of a multiplexed packet causes packet loss for all of the flows
   within the multiplexed packet.

   When the expected loss rate in a tunnel is relatively low (less than
   perhaps 5%), the robust operation (described in [ECRTP]) should be
   sufficient to ensure delivery of state changes.  This robust
   operation is characterized by a parameter N which means that the
   probability of more than N adjacent packets getting lost on the
   tunnel is small.

   A value of N=1 will protect against the loss of a single packet
   within a compressed session at the expense of bandwidth.  A value of
   N=2 will protect against the loss of two packets in a row within a
   compressed session and so on.  Higher values of N have higher
   bandwidth penalties.

   The optimal value of N will depend on the loss rate in the tunnel.
   If the loss rate is high (above perhaps 5%) more advanced techniques
   must be employed.  Those techniques are beyond the scope of this
   document.

3.3.  Bandwidth calculation for Voice and Video Applications

   The following formula uses the factors described above to model per-
   flow bandwidth usage for both voice and video applications.  These
   variables are defined:

   SOV-TCRTP, unit: octet.  Per-payload overhead of ECRTP and the
          multiplexed PPP header.  This value does not include
          additional overhead for updating IP ID or the RTP Time Stamp
          fields (see [ECRTP] for details on IP ID).  The value assumes
          the use of the COMPRESSED_RTP payload type.  It consists of 1
          octet for the ECRTP context ID, 1 octet for COMPRESSED_RTP
          flags, 2 octets for the UDP checksum, 1 octet for PPP
          protocol ID, and 1 octet for the multiplexed PPP length
          field.  The total is 6 octets.

   POV-TCRTP, unit: octet.  Per-packet overhead of tunneled ECRTP.
          This is the overhead for the tunnel header and the
          multiplexed PPP payload type.  This value is 20 octets for

Koren, Thompson, Wing                                        [Page 10]


INTERNET-DRAFT    Tunneled Multiplexed Compressed RTP    August, 2004


          the IP header, 4 octets for the L2TPv3 header and 1 octet for
          the multiplexed PPP protocol ID.  The total is 25 octets.

   TRANSMIT-LENGTH, unit: milliseconds.  The average duration of a
          transmission (such as a talk spurt for voice streams).

   SOV-TSTAMP, unit: octet.  Additional per-payload overhead of the
          COMPRESSED_UDP header that includes the absolute time stamp
          field.  This value includes 1 octet for the extra flags field
          in the COMPRESSED_UDP header and 4 octets for the absolute
          time stamp for a total of 5 octets.

   SOV-IPID, unit: octet.  Additional per-payload overhead of the
          COMPRESSED_UDP header that includes the absolute IPID field.
          This value includes 2 octets for the absolute IPID.  This
          value also includes 1 octet for the extra flags field in the
          COMPRESSED_UDP header.  The total is 3 octets.

   IPID-RATIO, unit: integer values 0 or 1.  Indicates the frequency at
          which IPID will be updated by the compressor. If IPID is
          changing randomly and thus always needs to be updated, then
          the value is 1. If IPID is changing by a fixed constant
          amount between payloads of a flow, then IPID-RATIO will be 0.
          The value of this variable does not consider the IPID value
          at the beginning of a voice or video transmission, as that is
          considered by the variable TRANSMIT-LENGTH.  The
          implementation of the sending IP stack and RTP application
          controls this behavior.  See section 1.1.

   NREP, unit: integer (usually a number between 1 and 3).  This is the
          number of times an update field will be repeated in ECRTP
          headers to increase the delivery rate between the compressor
          and decompressor.  For this example, we will assume NREP=2.

   PAYLOAD-SIZE, unit: octets. The size of the RTP payload in octets.

   MUX-SIZE, unit: count.  The number of PPP payloads multiplexed into
          one multiplexed PPP payload.

   SAMPLE-PERIOD, unit: milliseconds.  The average delay between
          transmissions of voice or video payloads for each flow in the
          multiplex.  For example, in voice applications the value of
          this variable would be 10ms if all calls have a 10ms sample
          period.

   The formula is:

     SOV-TOTAL = SOV-TCRTP + SOV-TSTAMP * (NREP * SAMPLE-PERIOD /
               TRANSMIT-LENGTH) + SOV-IPID * IPID-RATIO

Koren, Thompson, Wing                                        [Page 11]


INTERNET-DRAFT    Tunneled Multiplexed Compressed RTP    August, 2004



     BANDWIDTH = ((PAYLOAD-SIZE + SOV-TOTAL + (POV-TCRTP / MUX-SIZE)) *
               8) / SAMPLE-PERIOD)

   The results are:

   BANDWIDTH, unit: kilobits per second.  The average amount of
             bandwidth used per voice or video flow.

   SOV-TOTAL = The total amount of per-payload overhead associated with
             tunneled ECRTP.  It includes the per-payload overhead of
             ECRTP and PPP, timestamp update overhead, and IPID update
             overhead.

3.3.1.    Voice Bandwidth Calculation Example

   To create an example for a voice application using the above
   formulas, we will assume the following usage scenario.  Compressed
   voice streams using G.729 compression with a 20 millisecond
   packetization period.  In this scenario, VAD is enabled and the
   average talk spurt length is 1500 milliseconds.  The IPID field is
   changing randomly between payloads of streams.  There is enough
   traffic in the tunnel to allow 3 multiplexed payloads.  The
   following values apply:

        SAMPLE-PERIOD      = 20 milliseconds
        TRANSMIT-LENGTH    = 1500 milliseconds
        IPID-RATIO         = 1
        PAYLOAD-SIZE       = 20 octets
        MUX-SIZE           = 3

   For this example, per call bandwidth is 16.4 kbits/sec.  Classical
   CRTP over a single HDLC link using the same factors as above yields
   12.4 kbits/sec.

   The effect of IPID can have a large effect on per call bandwidth.
   If the above example is recalculated using an IPID-RATIO of 0, then
   the per call bandwidth is reduced to 13.8 kbits/sec.  Classical CRTP
   over a single HDLC link using these same factors yields 11.2
   kbits/call.

3.3.2.    Voice Bandwidth Comparison Table

   Using 5 simultaneous calls, no voice activity detection (VAD), G.729
   with 20ms packetization interval, not considering RTCP overhead:

       Normal VoIP over PPP:            124kbps
       with classical CRTP on a link:    50kbps (savings: 59%)
       with TCRTP over PPP:              62kbps (savings: 50%)

Koren, Thompson, Wing                                        [Page 12]


INTERNET-DRAFT    Tunneled Multiplexed Compressed RTP    August, 2004


       with TCRTP over AAL5:             85kbps (savings: 31%)

3.3.3.    Video Bandwidth Calculation Example

   Since TCRTP can be used to save bandwidth on any type of RTP
   encapsulated flow, it can be used to save bandwidth for video
   applications.  This section documents an example of TCRTP based
   bandwidth savings for MPEG-2 encoded video.

   To create an example for a video application using the above
   formulas, we will assume the following usage scenario.  RTP
   encapsulation of MPEG System and Transport Streams is performed as
   described in RFC 2250.  Frames for MPEG-2 encoded video are sent
   continuously, so the TRANSMIT-LENGTH variable in the bandwidth
   formula is essentially infinite.  The IPID field is changing
   randomly between payloads of streams.  There is enough traffic in
   the tunnel to allow 3 multiplexed payloads.  The following values
   apply:

        SAMPLE-PERIOD      = 2.8 milliseconds
        TRANSMIT-LENGTH    = infinite
        IPID-RATIO         = 1
        PAYLOAD-SIZE       = 1316 octets
        MUX-SIZE           = 3

   For this example, per flow bandwidth is 3.8 Mbits/sec.  MPEG video
   with no header compression using the same factors as above yields
   3.9 Mbits/sec.  While TCRTP does provide some bandwidth savings for
   video, the ratio of transmission headers to payload is so small that
   the bandwidth savings are insignificant.

3.3.4.    TCRTP over ATM

   IP transport over AAL5 causes a quantizing effect to bandwidth
   utilization due to the packets always being multiples of ATM cells.

   For example, the payload size for G.729 using 10 millisecond
   packetization interval is 10 octets.  This is much smaller than the
   payload size of an ATM cell (48 octets).  When classical CRTP [CRTP]
   is used on a link-by-link basis, the IP overhead to payload ratio is
   quite good.  However, AAL5 encapsulation and its cell padding always
   force the minimum payload size to be one ATM cell, which results in
   poor bandwidth utilization.

   Instead of wasting this padding, the multiplexing of TCRTP allows
   this previously wasted space in the ATM cell to contain useful data.
   This is one of the main reasons why multiplexing has such a large
   effect on bandwidth utilization with Voice over IP over ATM.


Koren, Thompson, Wing                                        [Page 13]


INTERNET-DRAFT    Tunneled Multiplexed Compressed RTP    August, 2004


   This multiplexing efficiency of TCRTP is similar to AAL2 sub-cell
   multiplexing described in [I.363.1].  Unlike AAL2 sub-cell
   multiplexing, however, TCRTP's multiplexing efficiency isn't limited
   to only ATM networks.

3.3.5.    TCRTP over non-ATM networks

   When TCRTP is used with other layer 2 encapsulations that do not
   have a minimum PDU size, the benefit of multiplexing is not as
   great.

   Depending upon the exact overhead of the layer 2 encapsulation, the
   benefit of multiplexing might be slightly better or worse than link-
   by-link CRTP header compression.  The per-payload overhead of CRTP
   tunneling is either 4 or 6 octets.  If classical CRTP plus layer 2
   overhead is greater than this amount, TCRTP multiplexing will
   consume less bandwidth than classical CRTP when the outer IP header
   is amortized over a large number of payloads.

   The payload breakeven point can be determined by the following
   formula:

     POV-L2 * MUX-SIZE >= POV-L2 + POV-TUNNEL + POV-PPPMUX + SOV-PPPMUX
          * MUX-SIZE

   Where:

     POV-L2, unit: octet.  Layer 2 packet overhead: 5 octets for HDLC
          encapsulation

     POV-TUNNEL, unit: octet.  Packet overhead due to tunneling: 24
          octets IP header and L2TPv3 header

     POV-PPPMUX, unit: octet.  Packet overhead for the multiplexed PPP
          protocol ID: 1 octet

     SOV-PPPMUX, unit: octet.  Per-payload overhead of PPPMUX, which is
          comprised of the payload length field and the ECRTP protocol
          ID.  The value of SOV-PPPMUX is typically 1, 2, or 3.

   If using HDLC as the layer 2 protocol, the breakeven point using the
   above formula is when MUX-SIZE = 7.  Thus 7 voice or video flows
   need to be multiplexed to make TCRTP as bandwidth-efficient as lijnk
   by link CRTP ocmpression.

4.   Example implementation of TCRTP


Koren, Thompson, Wing                                        [Page 14]


INTERNET-DRAFT    Tunneled Multiplexed Compressed RTP    August, 2004


   This section describes an example implementation of TCRTP.
   Implementations of TCRTP may be done in many ways as long as the
   requirements of the associated RFCs are met.

   Here is the path an RTP packet takes in this implementation:

         +-------------------------------+             ^
         |          Application          |             |
         +-------------------------------+             |
         |              RTP              |             |
         +-------------------------------+        Application and
         |              UDP              |            IP stack
         +-------------------------------+             |
         |              IP               |             |
         +-------------------------------+             V
                         |
                         |  IP forwarding
                         |
         +-------------------------------+             ^
         |             ECRTP             |             |
         +-------------------------------+             |
         |            PPPMUX             |             |
         +-------------------------------+          Tunnel
         |             PPP               |         Interface
         +-------------------------------+             |
         |             L2TP              |             |
         +-------------------------------+             |
         |              IP               |             |
         +-------------------------------+             V
                         |
                         |  IP forwarding
                         |
         +-------------------------------+             ^
         |            Layer 2            |             |
         +-------------------------------+          Physical
         |            Physical           |          Interface
         +-------------------------------+             V

   A protocol stack is configured to create an L2TP tunnel interface to
   a destination host.  The tunnel is configured to negotiate the PPP
   connection (using NCP IPCP) with ECRTP header compression and
   PPPMUX.  IP forwarding is configured to route RTP packets to this
   tunnel.  The destination UDP port number could distinguish RTP
   packets from non- RTP packets.

   The transmitting application gathers the RTP data from one source,
   and formats an RTP packet. Lower level application layers add UDP
   and IP headers to form a complete IP packet.


Koren, Thompson, Wing                                        [Page 15]


INTERNET-DRAFT    Tunneled Multiplexed Compressed RTP    August, 2004


   The RTP packets are routed to the tunnel interface where headers are
   compressed, payloads multiplexed, and then tunneled to the
   destination host.

   The operation of the receiving node is the same as the transmitting
   node in reverse.

4.1.  Suggested PPP and L2TP negotiation for TCRTP

   This section describes the necessary PPP and LT2P negotiations
   necessary for establishing a PPP connection and L2TP tunnel with
   L2TP header compression.  The negotiation is between two peers:
   Peer1 and Peer2.

4.2.  PPP negotiation TCRTP

   The Point-to-Point Protocol is described in [PPP].

4.2.1.    LCP negotiation

   Link Control Processing (LCP) is described in [PPP].

4.2.1.1.    Link Establishment

              Peer1                       Peer2
              -----                       -----
     Configure-Request (no options) ->
                                     <- Configure-Ack
                                     <- Configure-Request (no options)
     Configure-Ack                  ->

4.2.1.2.    Link Tear Down

     Terminate-Request              ->
                                     <- Terminate-Ack

4.2.2.    IPCP negotiation

   The protocol exchange here is described in [IPHCOMP], [PPP], and
   [ECRTP].

              Peer1                       Peer2
              -----                       -----
     Configure-Request              ->
       Options:
       IP-Compression-Protocol
         Use protocol 0x61
         and sub-parameters
         as described in

Koren, Thompson, Wing                                        [Page 16]


INTERNET-DRAFT    Tunneled Multiplexed Compressed RTP    August, 2004


         [IPCP-HC] and [ECRTP]
                                     <- Configure-Ack
                                     <- Configure-Request
                                          Options:
                                          IP-Compression-Protocol
                                            Use protocol 0x61
                                            and sub-parameters
                                            as described in
                                            [IPCP-HC] and [ECRTP]
     Configure-Ack                  ->

4.3.  L2TP negotiation

   L2TP is described in [L2TPv3].

4.3.1.    Tunnel Establishment

              Peer1                       Peer2
              -----                       -----
     SCCRQ                          ->
       Mandatory AVP's:
       Message Type
       Protocol Version
       Host Name
       Framing Capabilities
       Assigned Tunnel ID
                                     <- SCCRP
                                          Mandatory AVP's:
                                          Message Type
                                          Protocol Version
                                          Host Name
                                          Framing Capabilities
                                          Assigned Tunnel ID
     SCCCN                          ->
     Mandatory AVP's:
       Message Type
                                     <- ZLB

4.3.2.    Session Establishment

              Peer1                       Peer2
              -----                       -----
     ICRQ                           ->
       Mandatory AVP's:
       Message Type
       Assigned Session ID
       Call Serial Number
                                         <- ICRP
                                          Mandatory AVP's:

Koren, Thompson, Wing                                        [Page 17]


INTERNET-DRAFT    Tunneled Multiplexed Compressed RTP    August, 2004



                                          Message Type
                                          Assigned Session ID
     ICCN                           ->
       Mandatory AVP's:
       Message Type
       Tx (Connect Speed)
       Framing Type
                                     <- ZLB

4.3.3.    Tunnel Tear Down

              Peer1                       Peer2
              -----                       -----
     StopCCN                        ->
       Mandatory AVP's:
       Message Type
       Assigned Tunnel ID
       Result Code
                                     <- ZLB

5.   IANA Considerations

   This document does not require any assignments from IANA.

6.   Security Considerations

   This document describes a method for combining several existing
   protocols implementing compression, multiplexing, and tunneling of
   RTP streams.  Attacks on the component technologies of TCRTP include
   attacks on RTP/RTCP headers and payloads carried within a TCRTP
   session, attacks on the compressed headers, attacks on the
   multiplexing layer, or attacks on the tunneling negotiation or
   transport.  The security issues associated individually with each of
   those component technologies are addressed in their respective
   specifications, [ECRTP], [PPP-MUX], [L2TPv3], along with the
   security considerations for RTP itself [RTP].

   However, there may be additional security considerations arising
   from the use of these component technologies together.  For example,
   there may be an increased risk of unintended misdelivery of packets
   from one stream in the multiplex to another due to a protocol
   malfunction or data error because the addressing information is more
   condensed.  This is particularly true if the tunnel is transmitted
   over a link-layer protocol that allows delivery of packets
   containing bit errors in combination with a tunnel transport layer
   option that does not checksum all of the payload.


Koren, Thompson, Wing                                        [Page 18]


INTERNET-DRAFT    Tunneled Multiplexed Compressed RTP    August, 2004


   The opportunity for malicious misdirection may be increased relative
   to that for a single RTP stream transported by itself because
   addressing information must be unencrypted for the header
   compression and multiplexing layers to function.

   The primary defense against misdelivery is to make the data unusable
   to unintended recipients through cryptographic techniques.  The
   basic method for encryption provided in the RTP specification [RTP]
   is not suitable because it encrypts the RTP and RTCP headers along
   with the payload.  However, the RTP specification also allows
   alternative approaches to be defined in separate profile or payload
   format specifications wherein only the payload portion of the packet
   would be encrypted so header compression may be applied to the
   encrypted packets.  One such profile [SRTP] provides more
   sophisticated and complete methods for encryption and message
   authentication than the basic approach in [RTP].  Additional methods
   may be developed in the future.  Appropriate cryptographic
   protection should be incorporated into all TCRTP applications.

7.   Acknowledgements

   The authors would like to thank the authors of RFC2508, Stephen
   Casner and Van Jacobson, and the authors of RFC2507, Mikael
   Degermark, Bjorn Nordgren, and Stephen Pink.

   The authors would also like to thank Dana Blair, Alex Tweedley,
   Paddy Ruddy, Francois Le Faucheur, Tim Gleeson, Matt Madison,
   Hussein Salama, Mallik Tatipamula, Mike Thomas, Mark Townsley,
   Andrew  Valencia, Herb Wildfeuer, J. Martin Borden, John
   Geevarghese, and Shoou Yiu.

8.   References

   Normative References

     [PPP-MUX] R. Pazhyannur, I. Ali, C. Fox, "PPP Multiplexing",
          RFC3153, August 2001.

     [ECRTP] T. Koren, S. Casner, J. Geevarghese, B. Thompson, P.
          Ruddy, " Enhanced Compressed RTP (CRTP) for Links with High
          Delay, Packet Loss and Reordering",RFC3545, July 2003.

     [CRTP] S. Casner, V. Jacobson, "Compressing IP/UDP/RTP Headers for
          Low-Speed Serial Links", RFC2508, February 1999.

     [IPHCOMP] M. Degermark, B. Nordgren, S. Pink, "IP Header
          Compression", RFC2507, February 1999.


Koren, Thompson, Wing                                        [Page 19]


INTERNET-DRAFT    Tunneled Multiplexed Compressed RTP    August, 2004


     [IPCP-HC] M. Engan, S. Casner, C. Bormann, "IP Header Compression
          over PPP", RFC2509, February 1999.

     [RTP] H. Schulzrinne, S. Casner, R. Frederick, V. Jacobson, "RTP:
          A Transport Protocol for Real-Time Applications", RFC1889,
          January 1996.

     [L2TP] M. Townsley, A. Valencia, A. Rubens, G. Pall, G. Zorn, B.
          Palter, "Layer Two Tunneling Protocol "L2TP"", RFC2661,
          August 1999.

     [L2TPv3] J. Lau, M. Townsley, I. Goyret, "Layer Two Tunneling
          Protocol (Version 3)", draft-ietf-l2tpext-l2tp-base-14.txt,
          June 2004.

     [I.363.2] ITU-T, "B-ISDN ATM Adaptation layer specification: Type
          2 AAL", I.363.2, September 1997.

     [EF-PHB] V. Jacobson, K. Nichols, K. Poduri, "An Expedited
          Forwarding PHB", RFC2598, June 1999.

     [PPP] W. Simpson, "The Point-to-Point Protocol (PPP)", RFC1661,
          July 1994.

   Informative References

     [SRTP] M. Baugher, D. McGrew, M. Naslund, E. Carrara, K. Norrman,
          "The Secure Real-time Transport Protocol (SRTP)",RFC3711,
          March 2004.

     [REORDER] G. Pelletier, L. Jonsson, K. Sandlund, "RObust Header
          Compression (ROHC): ROHC over Channels that can Reorder
          Packets", Work in Progress, <draft-pelletier-rohc-over-
          reordering-00.txt>, June 2004.

     [ROHC] Bormann, C., Burmeister, C., Degermark, M., Fukushima, H.,
          Hannu, H., Jonsson, L., Hakenberg, R., Koren, T., Le, K.,
          Liu, Z., Martensson, A., Miyazaki, A., Svanbro, K., Wiebke,
          T., Yoshimura, T. and H. Zheng, "RObust Header Compression
          (ROHC): Framework and four profiles: RTP, UDP, ESP, and
          uncompressed", RFC 3095, July 2001.


9.   Authors' Addresses

   Bruce Thompson
   170 West Tasman Drive
   San Jose, CA  95134-1706
   United States of America

Koren, Thompson, Wing                                        [Page 20]


INTERNET-DRAFT    Tunneled Multiplexed Compressed RTP    August, 2004



   Phone: +1 408 527 0446
   Email: brucet@cisco.com

   Tmima Koren
   170 West Tasman Drive
   San Jose, CA  95134-1706
   United States of America

   Phone: +1 408 527 6169
   Email: tmima@cisco.com


   Dan Wing
   170 West Tasman Drive
   San Jose, CA  95134-1706
   United States of America

   Email: dwing@cisco.com

10.  Copyright Notice

   Copyright (C) The Internet Society (2004).  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.

11.  Disclaimers

   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.


   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

Koren, Thompson, Wing                                        [Page 21]


INTERNET-DRAFT    Tunneled Multiplexed Compressed RTP    August, 2004


   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.



Koren, Thompson, Wing                                        [Page 22]