Monami6 Working Group                                           T. Ernst
Internet-Draft                                                     INRIA
Intended status: Informational
Expires: January 13, 2008                                   N. Montavont
Expires: April 26, 2007
                                                             R. Wakikawa
                                                         Keio University
                                                                   C. Ng
                                                Panasonic Singapore Labs
                                                          K. Kuladinithi
                                                    University of Bremen
                                                        October 23, 2006
                                                           July 12, 2007

   Motivations and Scenarios for Using Multiple Interfaces and Global

Status of this Memo

   By submitting this Internet-Draft, each author represents that any
   applicable patent or other IPR claims of which he or she is aware
   have been or will be disclosed, and any of which he or she becomes
   aware will be disclosed, in accordance with Section 6 of BCP 79.

   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-

   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."

   The list of current Internet-Drafts can be accessed at

   The list of Internet-Draft Shadow Directories can be accessed at

   This Internet-Draft will expire on April 26, 2007. January 13, 2008.

Copyright Notice

   Copyright (C) The Internet Society (2006). IETF Trust (2007).


   In this document, multihoming is investigated from a node an end-node point
   of view, and not from a site point of view as the term "multihoming"
   is commonly understood so far.  The purpose of this document is to
   explain the motivations for fixed and mobile nodes (hosts and
   routers) using multiple interfaces and the scenarios where this may
   end up using multiple global addresses on their interfaces.  Such
   multihoming configurations can bring a number of benefits once
   appropriate support mechanisms are put in place.  Interestingly, this
   analysis is generic, i.e. motivations and benefits of node
   multihoming apply to both fixed end nodes and mobile end nodes.

Table of Contents

   1.  Introduction . . . . . . . . . . . . . . . . . . . . . . . . .  4

   2.  Terminology  . . . . . . . . . . . . . . . . . . . . . . . . .  5

   3.  Scenarios and Motivations  . . . . . . . . . . . . . . . . . .  6
     3.1.  Need for Ubiquitous Access to the Internet . . . . . . . .  6
     3.2.  Need to Redirect Established Sessions  . . . . . . . . . .  6
     3.3.  Need to Set Up Preferences . . . . . . . . . . . . . . . .  6
     3.4.  Need to Select the Best Access Technology  . . . . . . . .  7
     3.5.  Need to Dispatch Traffic over Distinct Paths . . . . . . .  8
     3.6.  Need for Reliability . . . . . . . . . . . . . . . . . . .  8
     3.7.  Need to Accelerate Transmission  . . . . . . . . . . . . .  8  9

   4.  Goals and Benefits of Multihoming  . . . . . . . . . . . . . . 10
     4.1.  Permanent and Ubiquitous Access  . . . . . . . . . . . . . 11
     4.2.  Reliability  . . . . . . . . . . . . . . . . . . . . . . . 11
     4.3.  Flow Redirection . . . . . . . . . . . . . . . . . . . . . 12
     4.4.  Load Sharing . . . . . . . . . . . . . . . . . . . . . . . 12
     4.5.  Load Balancing/Flow Distribution . . . . . . . . . . . . . 12
     4.6.  Preference Settings  . . . . . . . . . . . . . . . . . . . 12
     4.7.  Aggregate Bandwidth  . . . . . . . . . . . . . . . . . . . 12

   5.  Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
     5.1.  Case 1: One Interface, Multiple Prefixes . . . . . . . . . 13
     5.2.  Case 2: Several Interfaces . . . . . . . . . . . . . . . . 15

   6.  Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
     6.1.  <!--Source-->Address Selection . . . . . . . . . . . . . . 18
     6.2.  Failure Discovery and Recovery Delay . . . . . . . . . . . 18
     6.3.  Change of Traffic Characteristics  . . . . . . . . . . . . 19
     6.4.  Address Change  Transparency . . . . . . . . . . . . . . . . . . . . . . . 19
   7.  Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 20

   8.  Contributors . . . . . . . . . . . . . . . . . . . . . . . . . 20

   9.  Acknowledgments  . . . . . . . . . . . . . . . . . . . . . . . 20

   10. Informative References . . . . . . . . . . . . . . . . . . . . 20

   Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 22
   Intellectual Property and Copyright Statements . . . . . . . . . . 24

1.  Introduction

   New equipments shipped on the market now often integrate several
   access technologies (both wired and wireless).  The main purpose of
   this integration is to federate all means of communications in order
   to access the Internet ubiquitously (from everywhere and at any time)
   as no single technology can be expected to be deployed everywhere.
   Flows may thus be redirected from one interface to the other
   following the loss of connectivity or change of the network
   conditions.  Besides providing an Internet access of wide spread and
   reach, integrating several access technologies also allow to increase
   bandwidth availability or to select the the most appropriate
   technology according to the type of flow or choices of the user,
   since each network interface has different cost, performance,
   bandwidth, access range, and reliability.

   Once multiple accesses are offered, users may want to select the most
   appropriate set of network interface(s) depending on the network
   environment, particularly in wireless networks which are mutable and
   less reliable than wired networks.  Users may also want to select the
   most appropriate interface per communication type or to combine a set
   of interfaces to get sufficient bandwidth.

   The purpose of this document is to emphasize the goals and benefits
   of multihoming for fixed and mobile hosts and routers in a generic
   fashion, i.e. without focusing on issues pertaining to hosts, or
   routers, nor mobility.  This document is indeed completing other
   documents focusing on these latter issues: Issues pertaining to site
   multihoming in fixed networks are discussed in [1].  Mobility issues
   pertaining to mobile nodes and mobile networks are respectively
   discussed in companion drafts [2] and [3].  Our document is targetted
   to IPv6, although our analysis may be applicable to IPv4 as well.
   The readers may refer to [4] for a description of the problem
   specific to Mobile IPv4.

   This document is organized as follows: first, the terms used in the
   document are defined before illustrating the motivations by means of
   some scenarios in Section 3.  These scenarios are then used to
   emphasize the goals and benefits of multihoming in Section 4.  Next
   follows in Section 5 an analysis of the achievable goals in two
   multihoming configurations, i.e. when the node either has a single
   interface or when it has multiple interfaces.  Section 6 concludes
   this document with a number of generic issues that will have to be
   solved in order to effectively meet multihoming goals and benefits.

2.  Terminology

   This draft is based on the terminology defined in [5].  For the
   purpose of clarity, we remind the definition of interface.  Terms
   related to multihoming are not known to be defined in existing IETF


      A node's point of attachment to a link (as defined in [5])

   Multihomed Node

      A node (either a host or a router) is multihomed when it has
      several IPv6 addresses to choose between, i.e. in the following
      cases when it is either:

      *  multi-prefixed: multiple prefixes are advertised on the link(s)
         the node is attached to, or.

      *  multi-interfaced: the node has multiple interfaces to choose
         between, on the same link or not.

3.  Scenarios and Motivations

   The following real-life scenarios highlight the motivations for
   multihoming.  Each scenario usually yields more than one of the goals
   and benefits outlined in Section 4.  All scenarios focus on wireless
   technologies though no mobility management may be involved (one can
   use wireless access at office).

3.1.  Need for Ubiquitous Access to the Internet

   Mona is just getting out of a meeting with customers in a building.
   She calls her head office.  This audio communication is initiated via
   a private wireless local area network (WLAN) link realized over one
   of the available Wi-Fi hot-spots in the building.  This is going to
   be a long call and she must attend another meeting a few minutes
   drive from here.  She walks to a taxi stand, and boards a taxi.  The
   audio communication is automatically transferred to the public
   wireless metropolitan area network (WMAN) over the Wi-Max
   metropolitan network
   deployed in the city, deployed, with no interruption of the

   This scenario illustrates the need for a mobile user to use multiple types of access
   technologies in order to maintain ongoing communications when a user
   is moving out of the coverage area of a specific technology.

3.2.  Need to Redirect Established Sessions

   Oliver is on his way to work in the passenger lounge waiting at a train station. his train.  He uses this
   opportunity and the presence of a WLAN hot-spot to download the news
   from his favorite on-line news channel.  While Oliver is downloading
   the news, he receives a phone video call over a his wide area 3G cellular
   link.  Oliver decides to initiate a video flow for this
   communication.  The bandwidth and traversal delay of the wide area cellular
   link is not adequate for the video conference, high quality video-conference, so both flows
   (video/audio) are transferred to the WLAN link provided by the hot-
   spot.  This transfer occurs transparently and without affecting the
   other active flows.

   This scenario illustrates the need for a nomadic user to dynamically
   redirect flows from one type of access technology to another based on
   some user preferences or traffic requirements.

3.3.  Need to Set Up Preferences

   Nami works at home for a publishing company using her connection to
   the Internet via ADSL and a low-speed dial up connection, a public and
   unrealiable 802.11b WLAN from the street.  She is also subscribed to
   digital video broadcasting. street and her 3G cellular phone.
   Because the public WLAN is not secure, and the dial-up connection too
   slow, she downloads email from checks her company company's email using the ADSL link her 3G phone even though
   the WLAN service is free.  When she
   it is accessing her personal free
   web-mail account, she would then use the WLAN service. expensive.  She has
   noticed uses the 802.11b link WLAN service for non-confidential
   activities such as web-browsing and video-conferencing.  The dial-up
   connection is unstable during the day, so she chooses moslty used to send requests transmit her completed work securely and periodic refreshments for joining the digital
   video broadcasting via ADSL rather than the free WLAN services.
   synchronizing her file system.

   This scenario illustrates the need in a fixed environment to
   simultaneously use multiple access technologies and to select the
   most appropriate one according to user preferences.  No assumptions
   are made whether flows need to be redirected or not from one access
   technology to another.  These preferences can be dynamic, or as in dynamic (e.g. the example
   WLAN link is only used if the signal is good and there is no unusual
   latency) or configured once for each application. application (e.g. applications
   exchanging confidential data always over the most secure link).

3.4.  Need to Select the Best Access Technology

   Alice is a paramedic.  Her ambulance is called to the scene of a car
   accident.  She initiates a communication to a hospital via a wide
   area cellular link for the relay of low bit-rate live video from the
   site of the crash to assess the severity of the accident.  It is
   identified that one of the passengers has suffered a severe head
   injury.  Alice decides to consult a specialist via video
   conferencing.  This session is initiated from the specialist via the
   same wide area cellular link.  Meanwhile, Alice requests for the
   download of the patient medical records from the hospital servers.
   She later decides in mid-session that the
   The wide area cellular link is too slow for this download, and thus transfers so the
   download is transfered to the ambulance satellite link.  Even though
   this link provides a significantly faster bit rate it has a longer
   traversal delay and only down-link is available.  For this, only the
   down-stream of the download is transferred while up-stream proceeds
   over the wide area cellular link.  Connectivity with between the parametic
   and the ambulance is managed over a WLAN link between the paramedic and the ambulance. link.  Even though Alice has
   performed a partial hand-off for the transfer of the download
   down-stream down-
   stream to the satellite link, the upstream and the video conferencing
   session remains on the wide area cellular link.  This serves best the
   time constraint requirements of the real time communications.

   This scenario illustrates the need in a mobile environment for both
   ubiquitous access to the Internet using whatever available interface
   and the need to dispatch flows to particular access media according
   to traffic characteristics or preferences.  It also illustrates that
   flows can be directed to separate media downlink and uplink.  The
   fact that the actual connection to the Internet is maintained via the
   ambulance to which the paramedic is connected to via a WLAN link
   illustrates to need to express preferences on the path to be taken
   from a remote computer (i.e. a mobile router in the ambulance in this

3.5.  Need to Dispatch Traffic over Distinct Paths

   Max drives his car and constantly keeps some sort of Internet
   connectivity through one of the many available access technologies. technologies
   solely managed by a dedicated on-board unit (OBU).  Data are further
   transmitted to other on-board units.  His car navigator downloads
   road information from the Internet and his car-audio plays on-line
   audio streaming.  When his streaming while data collected by sensors is transmitted to the
   car passes an area
   where both high-data-rate Wi-Max manufacturer (e.g. consumption, engine pressure) and low-data rate cellular network safety data
   is exchanged between surrounding vehicles (e.g. geographic position,
   speed, brakes on/off, accident alerts).  Toll bills are available, it distributes load to the Wi-Max access paid
   automatically and the
   cellular network access. displayed on his navigation screen, while road sign
   transmit information (speed limitation, traffic lights).

   When his car passes an area where only a wide coverage-range cellular
   network is available, the connection is safety related sessions are maintained via the
   cellular network. network whereas infotainment data is buffered and
   transmitted over high data rate network access when one becomes
   available.  Toll bills and road sign data are transmitted over a
   dedicated radio interface, whereas data exchanged between vehicles is
   transmitted over a preferred media.  Time-critical safety sessions
   are always given priority.

   This scenario illustrates the applicability of multihoming in road
   transportation and emphasizes more particularly the need to save
   traffic transiting in a particular access network when there is a
   possibility to send data over an alternative route.  The availability
   of multiple access medium and the variety of on-board units
   illustrates a NEMO [6] scenario as currently considered in the CALM
   Architecture designed by ISO TC204 WG16 for Intelligent
   Transportations Systems (ITS).  The exchange of safety data
   illustrates the ongoing work of the car-to-car communication
   consortium (C2C-CC)

3.6.  Need for Reliability

   Dr. Ingrid

   Ingrid, a doctor, performs an operation via long-distance medical
   system.  She watches a patient in a battle field over the screen
   which delivers real-time images of the patient.  Sensors on her arms
   deliver her operational actions and a robot performs the actual
   operation in the battle field.  Since the operation is critical, the
   delivery of patient images and Dr. Ingrid's action is done by bi-
   casting from/to multiple interfaces bound to a distinct technology or
   distinct radio range.  So in case packets are delayed or one of the
   interface fails to maintain connectivity to the network, her distant
   operation can be continued.

   This scenario illustrates the need to use multiple access
   technologies in order to improve reliability upon failure of one of
   the access technologies.

3.7.  Need to Accelerate Transmission

   Roku is at the airport waiting to board the plane.  She receives a
   call from her husband.  This audio communication is received via a
   WLAN link realized over one of the available hot-spots.  She knows
   this is going to be a long flight and wishes to catch up on some
   work.  Roku uses a WLAN connection to download the necessary data.
   However, there is not enough time before boarding and she decides to
   accelerate the download.  Her notebook is equipped with an additional
   WLAN interface.  She decides to use this  This additional WLAN interface is then used to
   connect to another access point, and distribute the different download flows are
   distributed between the two wireless interfaces.

   This scenario illustrates the need to use multiple accesses to the
   Internet in order to accelerate the amount of data that could be
   transmitted over a period of time.

4.  Goals and Benefits of Multihoming

   The scenarios presented in the previous section are now used to
   highlight the goals and benefits of node multihoming.  The goals
   cannot really be distinguished from the benefits, but there are
   several situations where multihomed is either advisable or
   beneficial.  These benefits and goals listed here are by no means
   distinct and separate; most of them overlap with one another.  It is
   not the objective here to classify the benefits and goals into
   different non-overlapping consituents.  Instead the objective is to
   list the possible benefits and goals different people have in mind
   when deploying a multihomed node.

   Figure 1 summarizes which goal applies to the scenarios introduced in
   Section 3.  Note that all these goals and benefits apply to both
   fixed end nodes and mobile end nodes, though the scenarios may either
   focus on a fixed used (F), or nomadic usage (N), or a mobile usage
   (M).  Nomadic and mobile users are both on the move, while a fixed
   user doesn't physically move.  The difference between nomadic usages
   and mobile usages is that sessions are not required to be maintained
   when the access network is changed as a result of physical move
   within the topology.  No assumptions are made whether mobility
   support mechanims may be useful or not in any of the fixed, nomadic
   and mobile usages in order to maintain sessions.  This is out of
   scope of the present document.

   |                                   |       Scenarios       |   |
   |                                   +---+---+---+---+---+---+---+
   | Goals                             | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
   | Ubiquitous Access                 | o |   |   | o | o |   |   |
   | Flow Redirection                  | o | o | o | o | o |   |   |
   | Reliability                       |   |   | o | o | o | o |   |
   | Load Sharing                      |   |   |   |   | o |   |   |
   | Load Balancing                    |   |   | o | o |   |   | o |
   | Preference Settings               |   | o | o | o |   |   |   |
   | Aggregate Bandwidth               |   |   |   | o |   |   | o |
   | Usage: F=Fixed N=Nomadic M=Mobile | M | N | F | M | M | F | N |

                 Figure 1: Goals Applying to Each Scenario

4.1.  Permanent and Ubiquitous Access

   To provide an extended coverage area via distinct access

   Multiple interfaces bound to distinct technologies can be used to
   ensure a permanent connectivity is offered, anywhere, anytime, with

4.2.  Reliability

   To act upon failure at one point of attachment, i.e. the functions of
   a system component (e.g. interface, access network) are assumed by
   secondary system components when the primary component becomes
   unavailable (e.g. failure).  Connectivity is guaranteed as long as at
   least one connection to the Internet is maintained.

   A potential means is to duplicate network component, another is to
   duplicate a particular flow simultaneously through different routes.
   This minimizes packet loss typically for real-time communication and
   burst traffic.  It also minimizes delay of packet delivery caused by
   congestion and achieves more reliable real-time communication than
   single-casting.  For mobile computing, bi-casting avoids dropping
   packets when a mobile node changes its interface during communication

4.3.  Flow Redirection

   To be able to redirect flow flows from one interface, or one address to
   another one, without having to re-initiate the flow.  This can be due
   to preference changes or upon network failure.

4.4.  Load Sharing

   To spread network traffic load among several routes.  This is
   achieved when traffic load is distributed among different connections
   between the node and the Internet [7]. [8].

4.5.  Load Balancing/Flow Distribution

   To separate a flow between multiple points of attachment
   (simultaneously active or not) of a node, usually chosing the less
   loaded connection or according to preferences on the mapping between
   flows and interfaces.

4.6.  Preference Settings

   This goal is to provide the user, the application or the ISP the
   ability to choose the preferred transmission technology or access
   network based on cost, efficiency, policies, bandwidth requirement,
   delay, etc.

4.7.  Aggregate Bandwidth

   This goal is to provide the user or the application with more

   Bandwidth available to the user or the application may be limited by
   the underlying technology (e.g.  GSM has scarce bandwidth) or by some
   policies (e.g. monthly rate paid by the user).  Multiple interfaces
   connected to different links or ISPs can increase the total bandwidth
   available to the user or application.

5.  Analysis

   From the definition of a multihomed node it follows that a multihomed
   node has several IPv6 addresses to choose between.  In order to
   expose the goals and benefits to manage in managing multihomed nodes, we
   propose to distinguish two main cases: either the node has only one
   interface, or the node has several interfaces.  In the former case,
   the node is multihomed when multilpe prefixes are advertised on the
   link the node is attached to.  This distinction is important and
   sometimes subtle but the implications are important.

5.1.  Case 1: One Interface, Multiple Prefixes

   The single-interfaced node is multihomed when several prefixes are
   advertised on its interface.  The node must therefore configure
   several IPv6 addresses.

   A typical example is a node with a 802.11b interface, connected to an
   access point.  The access point is connected trough an Ethernet link
   to two access routers.  Each access router is configured to advertise
   distinct network prefixes by Router Advertisements on the link and
   can be used as default router.  Several reasons may lead to configure
   two access routers on the same link: for instance, the access points
   may be shared between different ISPs, or two access routers may be
   used for redundancy or load sharing purposes.  The node will then
   build two global IPv6 addresses on its interface.

   We now analyse which of the goals detailed in Section 4 can be met
   with this configuration.

   o  Ubiquitous Access: NO

      Ubiquitous access cannot be guaranteed when the node loses
      Internet connectivity through its sole interface (e.g. the node is
      going outside the coverage area of its access point).

   o  Flow redirection: YES

      The node might need to redirect a flow from one address to another
      for several reasons.  For example, if one of the IPv6 prefix
      becomes unavailable, flows using the corresponding address from this prefix need to
      could be redirected on an to the address using another prefix obtained on the other prefix.

   o  Reliability: MAYBE
      In case of failure of one IPv6 prefix, one of the address of the
      node will not be valid anymore.  Another available address built
      from other prefixes may allow the node to recover this sort of

      Bi-casting can be performed to ensure the delivery of packets on
      the node.  To do so, more than one IPv6 address must be used
      simultaneously for one flow.  Bi-casting would allow the node to
      seamlessly change the address used on the node.

   o  Load sharing: YES

      Load Sharing can be performed in the network, according to the
      address used by the node.  The choice of the address used by the
      node and the router selection can be influenced by the load sharing
      rules.  This mostly benefits the network side: if different access
      routers or routes can be used to forward the node's traffic, it will share the
      traffic load on will be shared in the network.

   o  Load balancing/Flow Distribution: NO

      Load balancing cannot be performed as when the node has only one

   o  Preferences: YES

      The source address can be chosen according to preferences set up
      by the user, or according to preferences set up in the network
      (such as with the default router preferences option introduced in
      Router Advertisement [8]), [9]), or by the ISP.

   o  Aggregated Bandwidth: MAYBE

      With only one interface connected to a link, the node generally
      will not be able to benefit from an increased aggregated bandwidth
      with multiple prefixes.  However, this benefit might be gained
      indirectly.  For instance, by alternating between different
      addresses, the total throughput may be higher (eg. due to load
      sharing).  Also, some web and file transfer servers limit transfer
      bandwidths based on the client's address.  By using different
      addresses to connect to the same server, the node may also see an
      increase in file transfer rate.

5.2.  Case 2: Several Interfaces

   In this case, the node may use its multiple interfaces either
   alternatively or simultaneously.  If used simultaneously, the node
   uses several IPv6 addresses at the same time (at least one address
   per interface, or several if several prefixes are announced on the
   link(s) it is connected to).  If used alternatively, the node may
   switch between its interfaces (e.g, one at a time), which is the case
   described above in Section 5.1.  In the paragraphs below, we assume
   that multiple interfaces are used simultaneously.  We also note that
   multiple interfaces can be connected to the same link as well as to
   different links.  These configurations will imply different issues.
   All these multihomed configurations may yield different benefits to
   the node.

   A typical example is a node with two interfaces, each one on a
   different technology (e.g. a WLAN 802.11b interface and a 3GPP GPRS
   interface), in order to benefit from a better coverage area and the
   characteristics of each technology.

   We now analyse how each of the goals listed in Section 4 can be met
   with such multihomed configuration:

   o  Ubiquitous Access: MAYBE

      It is easier to guarantee ubiquitous access when the node has
      multiple interfaces since distinct technologies may be available
      at a given time according to the location and administrative

   o  Flow redirection: YES

      In case of a change in user preferences, or a failure, flows might
      need to be redirected from one interface to another one.  Flows
      can be redirected individually or all flows attached to an
      interface might be redirected at once.

   o  Reliability: YES

      Two levels of redundancy can be seen in this case: either one
      address of one interface is not valid anymore (e.g. because the
      corresponding prefix is not advertised on the link), or the node
      loses its internet connectivity through one interface.  In the
      former case, another IPv6 address available on the interface would
      allow the node to switch addresses for on-going flows.  In the
      latter case, another connection to the internet through another
      interface would allow it to redirect on-going flow from the
      previous interface to the new one.  In either cases the node needs
      to change the IPv6 address for on-going sessions from the no
      longer valid address to one of the address available on the target
      interface.  The redirection will trigger a decision process to
      choose the best target interface to redirect the flow to.

      Bi-casting might be used to ensure the packets delivery on the
      node.  It would also allow seamless redirection between two
      addresses / interfaces with zero packets loss.  Bi-casting can be
      performed if several IPv6 addresses can be simultaneously used for
      one flow.  One entity between the CN (included) and the node
      (excluded) must duplicate the traffic to the destination node.

   o  Load Sharing: YES

      This benefit is mainly for the network side: if different access
      routers or routes can be used to forward traffic going into and
      out of the node, they can share the traffic load on the network.
      If the node uses several addresses at the same time for its on-
      going sessions, load sharing can be performed in the network.
      This goal can be a parameter that helps the source address

   o  Load balancing/Flow Distribution: YES

      Load balancing can be achieved on the node if several interfaces
      are used simultaneously.  Several interfaces can be used to spread
      the traffic load on the node.  This implies the choice of the IPv6
      address to use for each flow and the ability to choose a different
      address for each flow.

   o  Preferences: YES

      Interface and address selection is required.  The problem can be
      seen exactly as in the first case (the node has only one
      interface) if we consider that the interface preference is a
      parameter for the address selection.  Therefore in this case, the
      interface selection/preference is a supplementary parameter in the
      address selection algorithm.

   o  Aggregated Bandwidth: YES
      With multiple interfaces connected to different links, the node
      generally will be able to benefit from an increased aggregated

6.  Issues

   In this section, we attempt to list a number of generic issues that
   will have to be solved in order to meet the multihoming goals.

   Figure 2 summarizes which issues apply to each of the which case detailed in the
   previous section. section (availability of a single interface or multiple
   interfaces).  The sign '+', '-' or '=' indicated is indicates if the issue is more
   important, less important, or equally important to solve for the case
   under consideration

   |                                    |   Cases   |
   |                                    +-----+-----+
   | Issues                             | (1) | (2) |
   | Source Address Selection           | o = | o = |
   | Recovery Delay                     | o   | o + |
   | Change of e2e Path Characteristics | o - | o + |
   | Address Change Transparency                       | o + | o + |

            Figure 2: Issues and their Importance for Each Case

6.1.  <!--Source-->Address Selection

   The multihomed node has several addresses, which implies the
   appropriate address must be chosen when an IPv6 communication is
   established (e.g. when a TCP connection is opened).  An address
   selection mechanism is therefore needed.

   The choice of the address can be influenced by many parameters: user
   preferences, ingress filtering, preference flag in Router
   Advertisement, destination prefix, type of interface, link
   characteristics, etc.

6.2.  Failure Discovery and Recovery Delay

   A particular access to the Internet may become unavailable while it
   is being used.  The time needed for detecting an address has become
   invalid and the time to redirect communications to one of its other
   addresses is considered as critical.  Efficient failure detection and
   recovery mechanisms are therefore required.

   Note that transport sessions with multihoming capabilies such as SCTP
   [10] may be able to continue easily since SCTP has built-in
   transmission rate control mechanims to take into account the
   differences between two paths.

6.3.  Change of Traffic Characteristics

   The change of path for a specific session (e.g. due to change of
   interface) may cause changes on the end-to-end path characteristics
   (higher delay, different PMTU, etc).  This could have an impact on
   upper layer protocols such as transport protocols (particularly TCP)
   or applications that are sensitive to changes.

6.4.  Address Change  Transparency

   In some situations, it will be necessary to divert some or all of the
   sessions from one interface or prefix to another (e.g. due to loss of
   network connection or the access router becoming unreachable - this
   could be particularly frequent for mobile nodes).  With no support
   mechanism, an address change would cause on-going sessions using the
   invalid former address to terminate, and to be restarted using the
   new address.  To avoid this, the node needs a recovery mechanism allowing to redirect the
   redirection of all current communications to one of its the other IPv6 addresses.
   addresses is needed.

   In the case of a mobile node changing its point of attachment using
   the same interface, all flows must be redirected to the new location
   in order to maintain sessions.  A mobility management solution may be
   required, such as Mobile IPv6 [10] [11] for mobile hosts or NEMO Basic
   Support [11] [6] for mobile routers.  Additional mechanisms may be needed
   if the node was using several addresses on its old previous link, such as
   which flow flows shall be to redirect, redirected, which address must be associated
   with the new address(es).  The scalability of the operations involved
   in the redirection of flows may also be an issue, if we consider that
   the node had several addresses on the old previous link and several flows
   and/or correspondents.  Issues pertaining to Mobile IPv6 and NEMO
   Basic Support are explained in companion drafts [2] and [3]

7.  Conclusion

   In this document we show the concrete need for studied multihoming at the level of the end node.  We emphasized

   A node is multihomed in a situation where it has multiple addresses,
   usually due to the availability of multiple interfaces on the node,
   or the announcement of multiple prefixes on the link it is attached

   This fits a number of needs and goals brings a number of potential
   benefits.  The availability of having multiple interfaces at addresses allows the communicating end node.  Such interfaces
   could be used as one use of
   an alternate address as the replacement of the other (ubiquitous another (permanent and
   ubiquitous access to the Internet, reliability) or the transmission
   of multiple flows simultaneously (load over different routes (flow
   redirection, load sharing, load balancing/flow distribution,
   preference settings or aggregate bandwidth).

   Such goals are

   This study is motivated for both fixed nodes and mobile nodes, but
   needs prevail motivation prevails for mobile nodes (hosts and routers).  Goals  The
   benefits of multihoming can only be met achieved once some issues are
   solved.  Issues  Generic issues were outlined in the present document,
   whereas issues specific to mobile hosts and mobile routers are
   investigated in documents of the MONAMI6 associated documents [2] and
   NEMO working groups at the IETF. [3] and,

8.  Contributors

   This document is based on an earlier document to which Thomas Noel
   (ULP, Strasbourg) and EunKyoung Paik (SNU, Seoul) also participated contributed in
   addition to the authors listed in the present document.

9.  Acknowledgments

   We would like to extend our gratitude to Niko A. Fikouras, Ken
   Nagami, Pekka Savola, Hesham Soliman, and many others who had
   provided valuable comments to this document.

10.  Informative References

   [1]   Abley, J., Black, B., and V. Gill, "Goals for IPv6 Site-
         Multihoming Architectures", RFC 3582, August 2003.

   [2]   Montavont, N., Wakikawa, R., Ernst, T., Ng, C., and K.
         Kuladinithi, "Analysis of Multihoming in Mobile IPv6",
         draft-ietf-monami6-mipv6-analysis-03 (work in progress),
         June 2006.
         July 2007.

   [3]   Ng, C., Paik, E., Ernst, T., and M. C. Bagnulo, "Analysis of Multihoming
         in Network Mobility Support",
         draft-ietf-nemo-multihoming-issues-06 (work in progress),
         June 2006.

   [4]   Fikouras, N., "Mobile IPv4 Flow Mobility Problem Statement",
         draft-nomad-mip4-flow-mobility-pb-00.txt (work in progress),
         Feb 2004.

   [5]   Manner, J. and M. Kojo, "Mobility Related Terminology",
         RFC 3753, June 2004.

   [6]   Devarapalli, V., Wakikawa, R., Petrescu, A., and P. Thubert,
         "Network Mobility (NEMO) Basic Support Protocol", RFC 3963,
         January 2005.

   [7]   Malki, K. and H. Soliman, "Simultaneous Bindings for Mobile
         IPv6 Fast Handovers", draft-elmalki-mobileip-bicasting-v6-06
         (work in progress), July 2005.


   [8]   Hinden, R. and D. Thaler, "IPv6 Host to Router Load Sharing",
         draft-ietf-ipv6-host-load-sharing-04 (work in progress),
         June 2005.


   [9]   Draves, R. and D. Thaler, "Default Router Preferences and More-
         Specific Routes", RFC 4191, November 2005.


   [10]  Stewart, R., Xie, Q., Morneault, K., Sharp, C., Schwarzbauer,
         H., Taylor, T., Rytina, I., Kalla, M., Zhang, L., and V.
         Paxson, "Stream Control Transmission Protocol", RFC 2960,
         October 2000.


   [11]  Johnson, D., Perkins, C., and J. Arkko, "Mobility Support in
         IPv6", RFC 3775, June 2004.

   [11]  Devarapalli, V., Wakikawa, R., Petrescu, A., and P. Thubert,
         "Network Mobility (NEMO) Basic Support Protocol", RFC 3963,
         January 2005.

Authors' Addresses

   Thierry Ernst
   INRIA Rocquencourt
   Domaine de Voluceau B.P. 105
   Le Chesnay,   78153

   Phone: +33-1-39-63-59-30
   Fax:   +33-1-39-63-54-91

   Nicolas Montavont
   Ecole Nationale Superieure des telecommunications de Bretagne
   2, rue de la chataigneraie
   Cesson Sevigne  35576

   Phone: (+33) 2 99 12 70 23

   Ryuji Wakikawa
   Keio University
   Department of Environmental Information, Keio University.
   5322 Endo
   Fujisawa, Kanagawa  252-8520

   Phone: +81-466-49-1100
   Fax:   +81-466-49-1395
   Chan-Wah Ng
   Panasonic Singapore Laboratories Pte Ltd
   Blk 1022 Tai Seng Ave #06-3530
   Tai Seng Industrial Estate
   Singapore  534415

   Phone: +65 65505420

   Koojana Kuladinithi
   University of Bremen
   ComNets-ikom,University of Bremen.
   Otto-Hahn-Allee NW 1
   Bremen, Bremen  28359

   Phone: +49-421-218-8264
   Fax:   +49-421-218-3601

Full Copyright Statement

   Copyright (C) The Internet Society (2006). IETF Trust (2007).

   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

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

   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


   Funding for the RFC Editor function is provided by the IETF
   Administrative Support Activity (IASA).