Internet Engineering Task Force (IETF)                             J. Yi
Request for Comments: 8218                           Ecole Polytechnique
Category: Experimental                                        B. Parrein
ISSN: 2070-1721                                     University of Nantes
                                                             August 2017

   Multipath Extension for the Optimized Link State Routing Protocol
                           Version 2 (OLSRv2)

Abstract

   This document specifies a multipath extension for the Optimized Link
   State Routing Protocol version 2 (OLSRv2) to discover multiple
   disjoint paths for Mobile Ad Hoc Networks (MANETs).  Considering the
   characteristics of MANETs, especially the dynamic network topology,
   using multiple paths can increase aggregated throughput and improve
   the reliability by avoiding single route failures.  The
   interoperability with OLSRv2 is retained.

Status of This Memo

   This document is not an Internet Standards Track specification; it is
   published for examination, experimental implementation, and
   evaluation.

   This document defines an Experimental Protocol for the Internet
   community.  This document is a product of the Internet Engineering
   Task Force (IETF).  It represents the consensus of the IETF
   community.  It has received public review and has been approved for
   publication by the Internet Engineering Steering Group (IESG).  Not
   all documents approved by the IESG are a candidate for any level of
   Internet Standard; see Section 2 of RFC 7841.

   Information about the current status of this document, any errata,
   and how to provide feedback on it may be obtained at
   http://www.rfc-editor.org/info/rfc8218.

Copyright Notice

   Copyright (c) 2017 IETF Trust and the persons identified as the
   document authors.  All rights reserved.

   This document is subject to BCP 78 and the IETF Trust's Legal
   Provisions Relating to IETF Documents
   (http://trustee.ietf.org/license-info) in effect on the date of
   publication of this document.  Please review these documents
   carefully, as they describe your rights and restrictions with respect
   to this document.  Code Components extracted from this document must
   include Simplified BSD License text as described in Section 4.e of
   the Trust Legal Provisions and are provided without warranty as
   described in the Simplified BSD License.

Table of Contents

   1.  Introduction  . . . . . . . . . . . . . . . . . . . . . . . .   3
     1.1.  Motivation and Experiments to Be Conducted  . . . . . . .   3
   2.  Terminology . . . . . . . . . . . . . . . . . . . . . . . . .   5
   3.  Applicability Statement . . . . . . . . . . . . . . . . . . .   6
   4.  Protocol Overview and Functioning . . . . . . . . . . . . . .   7
   5.  Parameters and Constants  . . . . . . . . . . . . . . . . . .   8
     5.1.  Router Parameters . . . . . . . . . . . . . . . . . . . .   8
   6.  Packets and Messages  . . . . . . . . . . . . . . . . . . . .   9
     6.1.  HELLO and TC messages . . . . . . . . . . . . . . . . . .   9
       6.1.1.  SOURCE_ROUTE TLV  . . . . . . . . . . . . . . . . . .   9
     6.2.  Datagram  . . . . . . . . . . . . . . . . . . . . . . . .   9
       6.2.1.  Source Routing Header in IPv4 . . . . . . . . . . . .   9
       6.2.2.  Source Routing Header in IPv6 . . . . . . . . . . . .  10
   7.  Information Bases . . . . . . . . . . . . . . . . . . . . . .  10
     7.1.  SR-OLSRv2 Router Set  . . . . . . . . . . . . . . . . . .  10
     7.2.  Multipath Routing Set . . . . . . . . . . . . . . . . . .  10
   8.  Protocol Details  . . . . . . . . . . . . . . . . . . . . . .  11
     8.1.  HELLO and TC Message Generation . . . . . . . . . . . . .  11
     8.2.  HELLO and TC Message Processing . . . . . . . . . . . . .  12
     8.3.  MPR Selection . . . . . . . . . . . . . . . . . . . . . .  12
     8.4.  Datagram Processing at the MP-OLSRv2 Originator . . . . .  12
     8.5.  Multipath Calculation . . . . . . . . . . . . . . . . . .  14
       8.5.1.  Requirements of Multipath Calculation . . . . . . . .  14
       8.5.2.  Multipath Dijkstra Algorithm  . . . . . . . . . . . .  15
     8.6.  Multipath Routing Set Updates . . . . . . . . . . . . . .  16
     8.7.  Datagram Forwarding . . . . . . . . . . . . . . . . . . .  17
   9.  Configuration Parameters  . . . . . . . . . . . . . . . . . .  17
   10. Security Considerations . . . . . . . . . . . . . . . . . . .  18
   11. IANA Considerations . . . . . . . . . . . . . . . . . . . . .  19
     11.1.  Message TLV Types  . . . . . . . . . . . . . . . . . . .  19
   12. References  . . . . . . . . . . . . . . . . . . . . . . . . .  19
     12.1.  Normative References . . . . . . . . . . . . . . . . . .  19
     12.2.  Informative References . . . . . . . . . . . . . . . . .  20
   Appendix A.  Examples of Multipath Dijkstra Algorithm . . . . . .  22
   Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . .  23  24
   Authors' Addresses  . . . . . . . . . . . . . . . . . . . . . . .  24

1.  Introduction

   The Optimized Link State Routing Protocol version 2 (OLSRv2)
   [RFC7181] is a proactive link state protocol designed for use in
   Mobile Ad Hoc Networks (MANETs).  It generates routing messages
   periodically to create and maintain a Routing Set, which contains
   routing information to all the possible destinations in the routing
   domain.  For each destination, there exists a unique Routing Tuple,
   which indicates the next hop to reach the destination.

   This document specifies an extension of the OLSRv2 protocol [RFC7181]
   to provide multiple disjoint paths when appropriate for a source-
   destination pair.  Because of the characteristics of MANETs
   [RFC2501], especially the dynamic topology, having multiple paths is
   helpful for increasing network throughput, improving forwarding
   reliability, and load-balancing.

   Multipath OLSRv2 (MP-OLSRv2), specified in this document, uses the
   Multipath Dijkstra algorithm Algorithm by default to explore multiple disjoint
   paths from a source router to a destination router based on the
   topology information obtained through OLSRv2 and to forward the
   datagrams in a load-balancing manner using source routing.  MP-OLSRv2
   is designed to be interoperable with OLSRv2.

1.1.  Motivation and Experiments to Be Conducted

   This document is an experimental extension of OLSRv2 that can
   increase the data forwarding reliability in dynamic and high-load
   MANET scenarios by transmitting datagrams over multiple disjoint
   paths using source routing.  This mechanism is used because:

   o  Disjoint paths can avoid single route failures.

   o  Transmitting datagrams through parallel paths can increase
      aggregated throughput.

   o  Some scenarios may require that some routers must (or must not) be
      used.

   o  Having control of the paths at the source benefits the load-
      balancing and traffic engineering.

   o  An application of this extension is in combination with Forward
      Error Correction (FEC) coding applied across packets (erasure
      coding) [WPMC11].  Because the packet drops are normally bursty in
      a path (for example, due to route failure), erasure coding is less
      effective in single path routing protocols.  By providing multiple
      disjoint paths, the application of erasure coding with multipath
      protocol is more resilient to routing failures.

   In existing deployments, while running code and simulations have
   proven the interest of multipath extension for OLSRv2 in certain
   networks,
   networks [GIIS14][WCNC08][ADHOC11], more experiments and experiences
   are still needed to understand the effects of the protocol specified
   in this Experimental RFC.  The multipath extension for OLSRv2 is
   expected to be revised and documented as a Standards Track RFC once
   sufficient operational experience is obtained.  Other than general
   experiences, including the protocol specification and
   interoperability with base OLSRv2 implementations, experiences in the
   following aspects are highly appreciated:

   o  Optimal values for the number of multiple paths (NUMBER_OF_PATHS,
      see Section 5) to be used.  This depends on the network topology
      and router density.

   o  Optimal values used in the metric functions.  Metric functions are
      applied to increase the metric of used links and nodes so as to
      obtain disjoint paths.  What kind of disjointness is desired (node
      disjoint or link disjoint) may depend on the Layer 2 protocol used
      and can be achieved by applying different sets of metric
      functions.

   o  Use of different metric types.  This multipath extension can be
      used with metric types that meet the requirement of OLSRv2, such
      as [RFC7779].  The metric type used also has an impact on the
      choice of metric functions as indicated in the previous bullet
      point.

   o  The impact of partial topology information to multipath
      calculation.  OLSRv2 maintains a partial topology information base
      to reduce protocol overhead.  Experience has shown that multiple
      paths can be obtained even with such partial information; however,
      depending on the Multipoint Relay (MPR) selection algorithm used,
      the disjointness of the multiple paths might be impacted depending
      on the Multipoint Relay (MPR) selection algorithm used.

   o  Use of IPv6 loose source routing.  In the current specification,
      only strict source routing is used for IPv6 based on [RFC6554].
      In [IPv6-SRH], the use of the loose source routing is also
      proposed in IPv6.  In scenarios where the length of the source
      routing header is critical, the loose source routing can be
      considered.

   o  Optimal choice of "key" routers for loose source routing.  In some
      cases, loose source routing is used to reduce overhead or for
      interoperability with OLSRv2 routers.  Other than the basic rules
      defined in the following parts of this document, optimal choices
      of routers to put in the loose source routing header can be
      further studied.

   o  Different path-selection schedulers.  Depending on the application
      type and transport layer type, either a per-flow scheduler or per-
      datagram scheduler is applied.  By default, the traffic load
      should be equally distributed in multiple paths.  In some
      scenarios, weighted scheduling can be considered: for example, the
      paths with lower metrics (i.e., higher quality) can transfer more
      datagrams or flows compared to paths with higher metrics.

   o  The impacts of the delay variation due to multipath routing.
      [RFC2991] brings out some concerns of multipath routing,
      especially variable latencies when per-datagram scheduling is
      applied.  Although current experiment results show that multipath
      routing can reduce the jitter in dynamic scenarios, some transport
      protocols or applications may be sensitive to the datagram
      reordering.

   o  The disjoint multipath protocol has an interesting application
      with erasure coding, especially for services like video/audio
      streaming [WPMC11].  The combination of erasure coding mechanisms
      and this extension is thus encouraged.

   o  Different algorithms to obtain multiple paths, other than the
      default Multipath Dijkstra algorithm Algorithm introduced in Section 8.5.2
      of this specification.

   o  The use of multitopology information.  By using [RFC7722],
      multiple topologies using different metric types can be obtained.
      Although there is no work defining how this extension can make use
      of the multitopology information base yet, experimentation with
      the use of multiple metrics for building multiple paths is
      encouraged.

   Comments are solicited and should be addressed to the MANET working
   group's mailing list at manet@ietf.org and/or the authors.

2.  Terminology

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and
   "OPTIONAL" in this document are to be interpreted as described in BCP
   14 [RFC2119] [RFC8174] when, and only when, they appear in all
   capitals, as shown here.

   This document uses the terminology and notation defined in [RFC5444],
   [RFC6130], and [RFC7181].  Additionally, it defines the following
   terminology:

   OLSRv2 Routing Process:  A routing process based on [RFC7181],
      without multipath extension specified in this document.

   MP-OLSRv2 Routing Process:  A Multipath Routing Process based on this
      specification as an extension to [RFC7181].

   SR-OLSRv2 Routing Process:  An OLSRv2 Routing Process that supports
      Source Routing (SR) or an MP-OLSRv2 Routing Process.

3.  Applicability Statement

   As an extension of OLSRv2, this specification is applicable to MANETs
   for which OLSRv2 is applicable (see [RFC7181]).  It can operate on
   single or multiple interfaces to discover multiple disjoint paths
   from a source router to a destination router.  MP-OLSRv2 is designed
   for networks with dynamic topology to avoid single route failure.  It
   can also provide higher aggregated throughput and load-balancing.

   In a router supporting MP-OLSRv2, MP-OLSRv2 does not necessarily
   replace OLSRv2 completely.  The extension can be applied for certain
   applications that are suitable for multipath routing (mainly video or
   audio streams) based on information such as a Diffserv codepoint
   [RFC2474].

   Compared to OLSRv2, this extension does not introduce any new message
   type.  A new Message TLV Type is introduced to identify the routers
   that support forwarding based on the source routing header.  It is
   interoperable with OLSRv2 implementations that do not have this
   extension: as the MP-OLSRv2 uses source routing, in IPv4 networks the
   interoperability is achieved using loose source routing headers; in
   IPv6 networks, it is achieved by eliminating routers that do not
   support IPv6 strict source routing.

   MP-OLSRv2 supports two different but interoperable multipath
   calculation approaches: proactive and reactive.  In the proactive
   calculation, the paths to all the destinations are calculated before
   they are needed.  In the reactive calculation, only the paths to
   desired destination(s) are calculated on demand.  The proactive
   approach requires more computational resources than the reactive one.
   The reactive approach requires the IP forwarding plane to trigger the
   multipath calculation.

   MP-OLSRv2 forwards datagrams using the source routing header.  As
   there are multiple paths to each destination, MP-OLSRv2 requires the
   IP forwarding plane to be able to choose which source route to be put
   in the source routing header based on the path scheduler defined by
   MP-OLSRv2.  For IPv4 networks, implementation of loose source routing
   is required following [RFC0791].  For IPv6 networks, implementation
   of strict source routing is required following the source routing
   header generation and processing defined in [RFC6554].

4.  Protocol Overview and Functioning

   This specification uses OLSRv2 [RFC7181] to:

   o  Identify all the reachable routers in the network.

   o  Identify a sufficient subset of links in the networks so that
      routes can be calculated to all reachable destinations.

   o  Provide a Routing Set containing the shortest routes from this
      router to all destinations.

   In addition, the MP-OLSRv2 Routing Process identifies the routers
   that support source routing by adding a new Message TLV in HELLO and
   Topology Control (TC) messages.  Based on the above information
   acquired, every MP-OLSRv2 Routing Process is aware of a reduced
   topology map of the network and the routers supporting source
   routing.

   A Multipath Routing Set containing the multipath information is
   maintained.  It may be either proactively calculated or reactively
   calculated:

   o  In the proactive approach, multiple paths to all possible
      destinations are calculated and updated based on control message
      exchange.  The routes are thus available before they are actually
      needed.

   o  In the reactive approach, a multipath algorithm is invoked on
      demand, i.e., only when there is a datagram to be sent from the
      source to the destination and there is no available Routing Tuple
      in the Multipath Routing Set.  This requires the IP forwarding
      information base to trigger the multipath calculation specified in
      Section 8.5 when no Multipath Routing Tuple is available.  The
      reactive operation is local to the router and no additional
      exchange of routing control messages is required.  When the paths
      are being calculated, the datagrams SHOULD be buffered unless the
      router does not have enough memory.

   Routers in the same network may choose either proactive or reactive
   multipath calculation independently according to their computation
   resources.  The Multipath Dijkstra algorithm Algorithm (defined in Section 8.5)
   is introduced as the default algorithm to generate multiple disjoint
   paths from a source to a destination, and such information is kept in
   the Multipath Routing Set.

   The datagram is forwarded based on source routing.  When there is a
   datagram to be sent to a destination, the source router acquires a
   path from the Multipath Routing Set.  The path information is stored
   in the datagram header using the source routing header.

5.  Parameters and Constants

   In addition to the parameters and constants defined in [RFC7181],
   this specification uses the parameters and constants described in
   this section.

5.1.  Router Parameters

   NUMBER_OF_PATHS:  The number of paths desired by the router.

   MAX_SRC_HOPS:  The maximum number of hops allowed to be put in the
      source routing header.  A value set to 0 means there is no
      limitation on the maximum number of hops.  In an IPv6 network, it
      MUST be set to 0 because [RFC6554] supports only strict source
      routing.  All the intermediate routers MUST be included in the
      source routing header, which is a various number of hops.  In an
      IPv4 network, it MUST be strictly less than 11 and greater than 0
      due to the length limit of the IPv4 header.

   CUTOFF_RATIO:  The ratio that defines the maximum metric of a path
      compared to the shortest path kept in the OLSRv2 Routing Set.  For
      example, the metric to a destination is R_metric based on the
      Routing Set.  Then, the maximum metric allowed for a path is
      CUTOFF_RATIO * R_metric.  CUTOFF_RATIO MUST be greater than or
      equal to 1.  Setting the number low makes it less likely that
      additional paths will be found -- for example, setting it to 1
      will mean only equal length paths are considered.

   SR_TC_INTERVAL:  The maximum time between the transmission of two
      successive TC messages by an MP-OLSRv2 Routing Process.

   SR_HOLD_TIME:  The minimum value in the TLV with Type = VALIDITY_TIME
      included in TC messages generated based on SR_TC_INTERVAL.

6.  Packets and Messages

   This extension employs the routing control messages HELLO and TC as
   defined in OLSRv2 [RFC7181] to obtain network topology information.
   For the datagram to support source routing, a source routing header
   is added to each datagram routed by this extension.  Depending on the
   IP version used, the source routing header is defined in this
   section.

6.1.  HELLO and TC messages

   HELLO and TC messages used by the MP-OLSRv2 Routing Process use the
   same format as defined in [RFC7181].  In addition, a new Message TLV
   Type is defined to identify the originator of the HELLO or TC message
   that supports source-route forwarding.  The new Message TLV Type is
   introduced for enabling MP-OLSRv2 as an extension of OLSRv2: only the
   routers supporting source-route forwarding can be used in the source
   routing header of a datagram because adding a router that does not
   understand the source routing header will cause routing failure.

6.1.1.  SOURCE_ROUTE TLV

   The SOURCE_ROUTE TLV is a Message TLV signaling that the message is
   generated by a router that supports source-route forwarding.  It can
   be an MP-OLSRv2 Routing Process or an OLSRv2 Routing Process that
   supports source-route forwarding.

   Every HELLO or TC message generated by a MP-OLSRv2 Routing Process
   MUST have exactly one SOURCE_ROUTE TLV without value.

   Every HELLO or TC message generated by an OLSRv2 Routing Process MUST
   have exactly one SOURCE_ROUTE TLV, if the OLSRv2 Routing Process
   supports source-route forwarding, and be willing to join the source
   route generated by other MP-OLSRv2 Routing Processes.  The existence
   of SOURCE_ROUTE TLV MUST be consistent for a specific OLSRv2 Routing
   Process, i.e., either it adds SOURCE_ROUTE TLV to all its HELLO/TC
   messages or it does not add SOURCE_ROUTE TLV to any HELLO/TC
   messages.

6.2.  Datagram

6.2.1.  Source Routing Header in IPv4

   In IPv4 [RFC0791] networks, the MP-OLSRv2 Routing Process employs the
   loose source routing header, as defined in [RFC0791].  It exists as
   an option header with option class 0 and option number 3.

   The source route information is kept in the "route data" field of the
   loose source routing header.

6.2.2.  Source Routing Header in IPv6

   In IPv6 [RFC8200] networks, the MP-OLSRv2 Routing Process employs the
   source routing header, as defined in Section 3 of [RFC6554], with
   IPv6 Routing Type 3.

   The source route information is kept in the "Addresses" field of the
   routing header.

7.  Information Bases

   Each MP-OLSRv2 Routing Process maintains the information bases as
   defined in [RFC7181].  Additionally, a Multipath Information Base is
   used for this specification.  It includes the protocol sets as
   defined below.

7.1.  SR-OLSRv2 Router Set

   The SR-OLSRv2 Router Set records the routers that support source-
   route forwarding.  This includes routers that run the MP-OLSRv2
   Routing Process or the OLSRv2 Routing Process with source-route
   forwarding support.  The set consists of SR-OLSRv2 Router Tuples: Routing Tuple:

   (SR_addr, SR_time)

   where:

      SR_addr is the originator address of the router that supports
      source-route forwarding.

      SR_time is the time until which the SR-OLSRv2 Router Routing Tuple is
      considered valid.

7.2.  Multipath Routing Set

   The Multipath Routing Set records the full path information of
   different paths to the destination.  It consists of Multipath Routing
   Tuples:
   Tuple:

   (MR_dest_addr, MR_path_set)

   where:

      MR_dest_addr is the network address of the destination; it is
      either the network address of an interface of a destination router
      or the network address of an attached network.

      MP_path_set contains the multiple paths to the destination and it
      consists of a set of Path Tuples.

   Each Path Tuple is defined as:

   (PT_metric, PT_address[1], PT_address[2], ..., PT_address[n])

   where:

      PT_metric is the metric of the path to the destination, measured
      in LINK_METRIC_TYPE defined in [RFC7181].

      PT_address[1, ..., n-1] are the addresses of intermediate routers
      to be visited, numbered from 1 to n-1, where n is the number of
      routers in the path, i.e., the hop count.

8.  Protocol Details

   This protocol is based on OLSRv2 and is extended to discover multiple
   disjoint paths from a source router to a destination router.  It
   retains the formats of the basic routing control packets and the
   processing of OLSRv2 to obtain the topology information of the
   network.  The main differences from the OLSRv2 Routing Process are
   the datagram processing at the source router and datagram forwarding.

8.1.  HELLO and TC Message Generation

   HELLO messages are generated according to Section 15.1 of [RFC7181],
   plus a single message TLV with Type := SOURCE_ROUTE included.

   TC messages are generated according to Section 16.1 of [RFC7181],
   plus a single message TLV with Type := SOURCE_ROUTE included.

   For the routers that do not generate TC messages according to
   [RFC7181], at least one TC message MUST be generated by an MP-OLSRv2
   Routing Process during the SR_TC_INTERVAL (Section 5), which MUST be
   greater than or equal to TC_INTERVAL.  Those TC messages MUST NOT
   carry any advertised neighbor addresses.  This serves for those
   routers to advertise the SOURCE_ROUTE TLV so that the other routers
   can be aware of the routers that are source-route enabled so as to be
   used as destinations of multipath routing.  The validity time
   associated with the VALIDITY_TIME TLV in such TC messages equals
   SR_HOLD_TIME, which MUST be greater than the SR_TC_INTERVAL.  If the
   TC message carries an optional INTERVAL_TIME TLV, it MUST have a
   value encoding the SR_TC_INTERVAL.

8.2.  HELLO and TC Message Processing

   HELLO and TC messages are processed according to Sections 15.3 and
   16.3 of [RFC7181].

   In addition to the reasons specified in [RFC7181] for discarding a
   HELLO message or a TC message on reception, a HELLO or TC message
   received MUST be discarded if it has more than one Message TLV with
   Type := = SOURCE_ROUTE.

   For every HELLO or TC message received, if there is a Message TLV
   with Type := SOURCE_ROUTE, create or update (if the Tuple exists
   already) the SR-OLSR Router Routing Tuple with:

   o  SR_addr := originator address of the HELLO or TC message

   o  SR_time := current_time + validity time of the TC or HELLO message
      defined in [RFC7181].

8.3.  MPR Selection

   Each MP-OLSRv2 Routing Process selects routing MPRs and flooding MPRs
   following Section 18 of [RFC7181].  In a mixed network with
   OLSRv2-only routers, the following considerations apply when
   calculating MPRs:

   o  MP-OLSRv2 routers SHOULD be preferred as routing MPRs to increase
      the possibility of finding disjoint paths using MP-OLSRv2 routers.

   o  The number of routing MPRs that run the MP-OLSRv2 Routing Process
      MUST be equal to or greater than NUMBER_OF_PATHS if there are
      enough MP-OLSRv2 symmetric neighbors.  Otherwise, all the MP-
      OLSRv2 routers are selected as routing MPRs, except the routers
      with willingness WILL_NEVER.

8.4.  Datagram Processing at the MP-OLSRv2 Originator

   If datagrams without a source routing header need to be forwarded
   using multiple paths (for example, based on the information of a
   Diffserv codepoint [RFC2474]), the MP-OLSRv2 Routing Process will try
   to find the Multipath Routing Tuple where:

   o  MR_dest_addr = destination of the datagram
   If no matching Multipath Routing Tuple is found and the Multipath
   Routing Set is maintained proactively, it indicates that there is no
   multipath route available to the desired destination.  The datagram
   is forwarded following the OLSRv2 Routing Process.

   If no matching Multipath Routing Tuple is found and the Multipath
   Routing Set is maintained reactively, the multipath algorithm defined
   in Section 8.5 is invoked to calculate the Multipath Routing Tuple to
   the destination.  If the calculation does not return any Multipath
   Routing Tuple, the following steps are aborted and the datagram is
   forwarded following the OLSRv2 Routing Process.

   If a matching Multipath Routing Tuple is obtained, the Path Tuples of
   the Multipath Routing Tuple are applied to the datagrams using either
   per-flow or per-datagram scheduling, depending on the transport layer
   protocol and the application used.  By default, per-flow scheduling
   is used, especially for the transport protocols that are sensitive to
   reordering, such as TCP.  The path-selection decision is made on the
   first datagram and all subsequent datagrams of the same flow use the
   same path.  If the path is detected broken breaks before the flow is closed, another
   path with the most similar metric is used.  Per-datagram scheduling
   is recommended if the traffic is insensitive to reordering such as
   unreliable transmission of media traffic or when erasure coding is
   applied.  In such a case, each datagram selects its paths
   independently.

   By default, the traffic load should be equally distributed in
   multiple paths.  Other path-scheduling mechanisms (e.g., assigning
   more traffic over better paths) are also possible and will not impact
   the interoperability of different implementations.

   The addresses in PT_address[1, ..., n-1] of the chosen Path Tuple are
   thus added to the datagram header as the source routing header.  For
   IPv6 networks, strict source routing is used; thus, all the
   intermediate routers in the path are stored in the source routing
   header following the format defined in Section 3 of [RFC6554] with
   the Routing Type set to 3.

   For IPv4 networks, loose source routing is used with the following
   rules:

   o  Only the addresses that exist in the SR-OLSR Router Set can be
      added to the source routing header.

   o  If the length of the path (n) is greater than MAX_SRC_HOPS
      (Section 5) or if adding the whole path information exceeds the
      MTU, only the "key" routers in the path are kept.  By default, the
      key routers are uniformly chosen in the path.  If further
      information, such as the capacity of the routers (e.g., battery
      life) or the routers' willingness in forwarding data, is
      available, the routers with higher capacity and willingness are
      preferred.

   o  The routers that are considered not appropriate for forwarding
      indicated by external policies should be avoided.

   It is not recommended to fragment the IP packet if the packet with
   the source routing header would exceed the minimum MTU along the
   path.  Depending on the size of the routing domain, the MTU should be
   at least 1280 + 40 (for the outer IP header) + 16 * diameter of the
   network in number of hops (for the source routing header).  If the
   links in the network have different MTU sizes, by using technologies
   like Path MTU Discovery, the routers are able to be aware of the MTU
   along the path.  The size of the datagram plus the size of IP headers
   (including the source routing header) should not exceed the minimum
   MTU along the path; otherwise, the source routing should not be used.

   If the destination of the datagrams is out of the MP-OLSRv2 routing
   domain, the datagram must be source routed to the gateway between the
   MP-OLSRv2 routing domain and the rest of the Internet.  The gateway
   MUST remove the source routing header before forwarding the datagram
   to the rest of the Internet.

8.5.  Multipath Calculation

8.5.1.  Requirements of Multipath Calculation

   The Multipath Routing Set maintains the information of multiple paths
   to the destination.  The Path Tuples of the Multipath Routing Set
   (Section 7.2) are generated based on a multipath algorithm.

   For each path to a destination, the algorithm must provide:

   o  The metric of the path to the destination,

   o  The list of intermediate routers on the path.

   For IPv6 networks, as strict source routing is used, only the routers
   that exist in the SR-OLSRv2 Router Set are considered in the path
   calculation, i.e., only the source-routing-supported routers can
   exist in the path.

   After the calculation of multiple paths, the metric of paths (denoted
   c_i for path i) to the destination is compared to the R_metric of the
   OLSRv2 Routing Tuple ([RFC7181]) to the same destination.  If the
   metric c_i is greater than R_metric * CUTOFF_RATIO (Section 5), the
   corresponding path i SHOULD NOT be used.  If less than two paths are
   found with metrics less than R_metric * CUTOFF_RATIO, the router
   SHOULD fall back to OLSRv2 Routing Process without using multipath
   routing.  This can happen if there are too many OLSRv2-only routers
   in the network, and requiring multipath routing may result in
   inferior paths.

   By invoking the multipath algorithm, up to NUMBER_OF_PATHS paths are
   obtained and added to the Multipath Routing Set by creating a
   Multipath Routing Tuple with:

   o  MR_dest_addr := destination of the datagram.

   o  An MP_path_set with calculated Path Tuples.  Each Path Tuple
      corresponds to a path obtained in the Multipath Dijkstra
      algorithm,
      Algorithm, with PT_metric := metric of the calculated path and
      PT_address[1, ..., n-1] := list of intermediate routers.

8.5.2.  Multipath Dijkstra Algorithm

   This section introduces the Multipath Dijkstra Algorithm as a default
   algorithm.  It tries to obtain disjoint paths when appropriate, but
   it does not guarantee strict disjoint paths.  The use of other
   algorithms is not prohibited, as long as the requirements described
   in Section 8.5.1 are met.  Using different multipath algorithms will
   not impact the interoperability.

   The general principle of the Multipath Dijkstra Algorithm [ADHOC11]
   is to use the Dijkstra algorithm Algorithm for multiple iterations and to look
   for the shortest path P[i] to the destination d at iteration i.
   After each iteration, the metric of used links is increased.
   Compared to the original Dijkstra's algorithm, the main modification
   consists in adding two incremental functions, named metric functions
   fp and fe, in order to prevent the next steps resulting in similar
   paths:

   o  fp(c) is used to increase metrics of arcs belonging to the
      previous path P[i-1] (with i>1), where c is the value of the
      previous metric.  This encourages future paths to use different
      arcs but not different vertices.

   o  fe(c) is used to increase metrics of the arcs that lead to
      intermediate vertices of the previous path P[i-1] (with i>1),
      where c is the value of the previous metric.  The "lead to" means
      that only one vertex of the arc belongs to the previous path
      P[i-1] while the other vertex does not.  The "intermediate" means
      that the source and destination vertices are not considered.

   Consider the simple example in Figure 1: a path P[i] S--A--D is
   obtained at step i.  For the next step, the metric of link S--A and A
   --D are to be increased using fp(c) because they belong to the path
   P[i].  A--B is to be increased using fe(c) because A is an
   intermediate vertex of path P[i], and B is not part of P[i].  B--D is
   unchanged.

                                          B
                                       /    \
                                      /      \
                                     /        \
                          S---------A-----------D

                                 Figure 1

   It is possible to choose a different fp and fe to get link-disjoint
   paths or node-disjoint paths as desired.  A recommendation for
   configuration of fp and fe is given in Section 9.

   To get NUMBER_OF_PATHS different paths, for each path P[i] (i = 1,
   ..., NUMBER_OF_PATHS):

   1.  Run Dijkstra's algorithm to get the shortest path P[i] for the
       destination d.

   2.  Apply metric function fp to the metric of links (in both
       directions) in P[i].

   3.  Apply metric function fe to the metric of links (in both
       directions) that lead to routers used in P[i].

   A simple example of the Multipath Dijkstra Algorithm is illustrated
   in Appendix A.

8.6.  Multipath Routing Set Updates

   The Multipath Routing Set MUST be updated when the Local Information
   Base, the Neighborhood Information Base, or the Topology Information
   Base indicate a change (including a change of any potentially used
   outgoing neighbor metric values) of the known symmetric links and/or
   attached networks in the MANET, hence, changing the Topology Graph as
   described in Section 17.7 of [RFC7181].  How the Multipath Routing
   Set is updated depends on whether the set is maintained reactively or
   proactively:

   o  In reactive mode, all the Tuples in the Multipath Routing Set are
      removed.  The new arriving datagrams will be processed as
      specified in Section 8.4.

   o  In proactive mode, the routes to all the destinations are updated
      according to Section 8.5.

8.7.  Datagram Forwarding

   In IPv4 networks, datagrams are forwarded using loose source routing
   as specified in Section 3.1 of [RFC0791].

   In IPv6 networks, datagrams are forwarded using strict source routing
   as specified in Section 4.2 of [RFC6554], except the applied routers
   are MP-OLSRv2 routers rather than RPL routers.  The last hop of the
   source route MUST remove the source routing header.

9.  Configuration Parameters

   This section gives default values and guidelines for setting
   parameters defined in Section 5.  Network administrators may wish to
   change certain or all the parameters for different network scenarios.
   As an experimental protocol, the users of this protocol are also
   encouraged to explore different parameter settings in various network
   environments and provide feedback.

   o  NUMBER_OF_PATHS := 3.  This parameter defines the number of
      parallel paths used in datagram forwarding.  Setting it to 1 makes
      the specification identical to OLSRv2.  Setting it to too large of
      a value may lead to unnecessary computational overhead and
      inferior paths.

   o  MAX_SRC_HOPS := 10, for IPv4 networks.  For IPv6 networks, it MUST
      be set to 0, i.e., no constraint on the maximum number of hops.

   o  CUTOFF_RATIO := 1.5.  It MUST be greater than or equal to 1.

   o  SR_TC_INTERVAL := 10 x TC_INTERVAL.  It MUST be greater than or
      equal to TC_INTERVAL.  It SHOULD be significantly greater than
      TC_INTERVAL to reduce unnecessary TC message generations.

   o  SR_HOLD_TIME := 3 x SR_TC_INTERVAL.  It MUST be greater than
      SR_TC_INTERVAL and SHOULD allow for a small number of lost
      messages.

   If the Multipath Dijkstra Algorithm is applied:

   o  fp(c) := 4*c, where c is the original metric of the link.

   o  fe(c) := 2*c, where c is the original metric of the link.

   The setting of metric functions fp and fc defines the preference of
   obtained multiple disjoint paths.  If id is the identity function,
   i.e., fp(c)=c, three cases are possible:

   o  if id=fe<fp, only increase the metric of related links;

   o  if id<fe=fp, apply equal increase to the metric of related nodes
      and links;

   o  if id<fe<fp, apply greater increase to the metric of related
      links.

   Increasing the metric of related links or nodes means avoiding the
   use of such links or nodes in the next path to be calculated.

10.  Security Considerations

   As an extension of [RFC7181], the security considerations and
   security architecture illustrated in [RFC7181] are applicable to this
   MP-OLSRv2 specification.  The implementations without security
   mechanisms are vulnerable to threats discussed in [RFC8116].

   In a mixed network with OLSRv2-only routers, a compromised router can
   add SOURCE_ROUTE TLVs in its TC and HELLO messages, which will make
   other MP-OLSRv2 Routing Processes believe that it supports source
   routing.  This will increase the possibility of being chosen as MPRs
   and put into the source routing header.  The former will make it
   possible to manipulate the flooding of TC messages and the latter
   will make the datagram pass through the compromised router.

   As with [RFC7181], a conformant implementation of MP-OLSRv2 MUST, at
   minimum, implement the security mechanisms specified in [RFC7183] to
   provide integrity and replay protection of routing control messages.

   The MP-OLSRv2 Routing Process MUST drop datagrams entering or exiting
   an OLSRv2/MP-OLSRv2 routing domain that contain a source routing
   header.  Compared to OLSRv2, the use of the source routing header in
   this specification introduces vulnerabilities related to source
   routing attacks, which include bypassing filtering devices, bandwidth
   exhaustion of certain routers, etc.  Those attacks are discussed in
   Section 5 of [RFC6554] and [RFC5095].  The influence is limited to
   the OLSRv2/MP-OLSRv2 routing domain because the source routing header
   is used only in the current routing domain.

   If the multiple paths are calculated reactively, the datagrams SHOULD
   be buffered while the paths are being calculated.  Because the path
   calculation is local and no control message is exchanged, the
   buffering time should be trivial.  However, depending on the CPU
   power and memory of the router, a maximum buffer size SHOULD be set
   to avoid occupying too much memory of the router.  When the buffer is
   full, the oldest datagrams are dropped.  A possible attack that a
   malicious application could launch would be one in which it initiates
   a large amount of datagrams to all the other routers in the network,
   thus triggering path calculation to all the other routers and during
   which the datagrams are buffered.  This might flush other legitimate
   datagrams.  But the impact of the attack is transient: once the path
   calculation is finished, the datagrams are forwarded and the buffer
   goes back to empty.

11.  IANA Considerations

   This section adds one new Message TLV, allocated as a new Type
   Extension to an existing Message TLV.

11.1.  Message TLV Types

   This specification updates the IANA registry "Type 7 Message TLV Type
   Extensions" by adding the new Type Extension SOURCE_ROUTE, as
   illustrated in Table 1.

   +-----------+--------------+------------------------+---------------+
   |    Type   |     Name     |      Description       | Reference     |
   | Extension |              |                        |               |
   +-----------+--------------+------------------------+---------------+
   |     2     | SOURCE_ROUTE |   Indicates that the   | This          |
   |           |              |   originator of the    | specification |
   |           |              |    message supports    |               |
   |           |              |      source-route      |               |
   |           |              | forwarding. No value.  |               |
   +-----------+--------------+------------------------+---------------+

     Table 1: SOURCE_ROUTE Type type for RFC 5444 Type 7 Message TLV Type Extensions
                                 [RFC5444]

12.  References

12.1.  Normative References

   [RFC0791]  Postel, J., "Internet Protocol", STD 5, RFC 791,
              DOI 10.17487/RFC0791, September 1981,
              <http://www.rfc-editor.org/info/rfc791>.

   [RFC2119]  Bradner, S., "Key words for use in RFCs to Indicate
              Requirement Levels", BCP 14, RFC 2119,
              DOI 10.17487/RFC2119, March 1997,
              <http://www.rfc-editor.org/info/rfc2119>.

   [RFC5444]  Clausen, T., Dearlove, C., Dean, J., and C. Adjih,
              "Generalized Mobile Ad Hoc Network (MANET) Packet/Message
              Format", RFC 5444, DOI 10.17487/RFC5444, February 2009,
              <http://www.rfc-editor.org/info/rfc5444>.

   [RFC6130]  Clausen, T., Dearlove, C., and J. Dean, "Mobile Ad Hoc
              Network (MANET) Neighborhood Discovery Protocol (NHDP)",
              RFC 6130, DOI 10.17487/RFC6130, April 2011,
              <http://www.rfc-editor.org/info/rfc6130>.

   [RFC6554]  Hui, J., Vasseur, JP., Culler, D., and V. Manral, "An IPv6
              Routing Header for Source Routes with the Routing Protocol
              for Low-Power and Lossy Networks (RPL)", RFC 6554,
              DOI 10.17487/RFC6554, March 2012,
              <http://www.rfc-editor.org/info/rfc6554>.

   [RFC7181]  Clausen, T., Dearlove, C., Jacquet, P., and U. Herberg,
              "The Optimized Link State Routing Protocol Version 2",
              RFC 7181, DOI 10.17487/RFC7181, April 2014,
              <http://www.rfc-editor.org/info/rfc7181>.

   [RFC7183]  Herberg, U., Dearlove, C., and T. Clausen, "Integrity
              Protection for the Neighborhood Discovery Protocol (NHDP)
              and Optimized Link State Routing Protocol Version 2
              (OLSRv2)", RFC 7183, DOI 10.17487/RFC7183, April 2014,
              <http://www.rfc-editor.org/info/rfc7183>.

   [RFC8174]  Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC
              2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174,
              May 2017, <http://www.rfc-editor.org/info/rfc8174>.

12.2.  Informative References

   [ADHOC11]  Yi, J., Adnane, A., David, S., and B. Parrein, "Multipath
              optimized link state routing for mobile ad hoc networks",
              Elsevier Ad Hoc Networks, Volume 9, Number 1, pp 28-47,
              DOI 10.1016/j.adhoc.2010.04.007, January 2011.

   [GIIS14]   Macedo, R., Melo, R., Santos, A., and M. Nogueria,
              "Experimental performance comparison of single-path and
              multipath routing in VANETs", In the Global Information
              Infrastructure and Networking Symposium (GIIS), Volume 1,
              Number 6, pp 15-19, DOI 10.1109/GIIS.2014.6934283,
              September 2014.

   [IPv6-SRH]
              Previdi, S., Ed., Filsfils, C., Raza, K., Leddy, J.,
              Field, B., Voyer, D., Bernier, S., Matsushima, S., Leung,
              I., Linkova, J., Aries, E., Kosugi, T., Vyncke, E.,
              Lebrun, D., Steinberg, D., and R. Raszuk, "IPv6 Segment
              Routing Header (SRH)", Work in Progress, draft-ietf-6man-
              segment-routing-header-07, July 2017.

   [RFC2474]  Nichols, K., Blake, S., Baker, F., and D. Black,
              "Definition of the Differentiated Services Field (DS
              Field) in the IPv4 and IPv6 Headers", RFC 2474,
              DOI 10.17487/RFC2474, December 1998,
              <http://www.rfc-editor.org/info/rfc2474>.

   [RFC2501]  Corson, S. and J. Macker, "Mobile Ad hoc Networking
              (MANET): Routing Protocol Performance Issues and
              Evaluation Considerations", RFC 2501,
              DOI 10.17487/RFC2501, January 1999,
              <http://www.rfc-editor.org/info/rfc2501>.

   [RFC2991]  Thaler, D. and C. Hopps, "Multipath Issues in Unicast and
              Multicast Next-Hop Selection", RFC 2991,
              DOI 10.17487/RFC2991, November 2000,
              <http://www.rfc-editor.org/info/rfc2991>.

   [RFC5095]  Abley, J., Savola, P., and G. Neville-Neil, "Deprecation
              of Type 0 Routing Headers in IPv6", RFC 5095,
              DOI 10.17487/RFC5095, December 2007,
              <http://www.rfc-editor.org/info/rfc5095>.

   [RFC7722]  Dearlove, C. and T. Clausen, "Multi-Topology Extension for
              the Optimized Link State Routing Protocol Version 2
              (OLSRv2)", RFC 7722, DOI 10.17487/RFC7722, December 2015,
              <http://www.rfc-editor.org/info/rfc7722>.

   [RFC7779]  Rogge, H. and E. Baccelli, "Directional Airtime Metric
              Based on Packet Sequence Numbers for Optimized Link State
              Routing Version 2 (OLSRv2)", RFC 7779,
              DOI 10.17487/RFC7779, April 2016,
              <http://www.rfc-editor.org/info/rfc7779>.

   [RFC8116]  Clausen, T., Herberg, U., and J. Yi, "Security Threats to
              the Optimized Link State Routing Protocol Version 2
              (OLSRv2)", RFC 8116, DOI 10.17487/RFC8116, May 2017,
              <http://www.rfc-editor.org/info/rfc8116>.

   [RFC8200]  Deering, S. and R. Hinden, "Internet Protocol, Version 6
              (IPv6) Specification", STD 86, RFC 8200,
              DOI 10.17487/RFC8200, July 2017,
              <http://www.rfc-editor.org/info/rfc8200>.

   [WCNC08]   Yi, J., Cizeron, E., Hamma, S., and B. Parrein,
              "Simulation and Performance Analysis of MP-OLSR for Mobile
              Ad hoc Networks", In Proceedings of the IEEE Wireless
              Communications and Networking Conference (WCNC),
              DOI 10.1109/WCNC.2008.395, 2008.

   [WPMC11]   Yi, J., Parrein, B., and D. Radu, "Multipath Routing
              Protocol for MANET: Application to H.264/SVC Video Content
              Delivery", Proceedings of the 14th International Symposium
              on Wireless Personal Multimedia Communications, 2011.

Appendix A.  Examples of Multipath Dijkstra Algorithm

   This appendix gives two examples of the Multipath Dijkstra algorithm. Algorithm.

   A network topology is depicted in Figure 2.

                              .-----A-----(2)
                             (1)   / \     \
                            /     /   \     \
                           S     (2)   (1)   D
                            \   /       \   /
                           (1) /         \ / (2)
                              B----(3)----C

                                 Figure 2

   The capital letters are the names of routers.  An arbitrary metric
   with value between 1 and 3 is used.  The initial metrics of all the
   links are indicated in the parentheses.  The incremental functions
   fp(c)=4c and fe(c)=2c are used in this example.  Two paths from
   router S to router D are demanded.

   On the first run of the Dijkstra algorithm, Algorithm, the shortest path S->A->D
   with metric 3 is obtained.

   The incremental function fp is applied to increase the metric of the
   link S-A and A-D, and fe is applied to increase the metric of the
   link A-B and A-C.  Figure 3 shows the link metrics after the
   increment.

                              .-----A-----(8)
                             (4)   / \     \
                            /     /   \     \
                           S     (4)   (2)   D
                            \   /       \   /
                           (1) /         \ / (2)
                              B----(3)----C

                                 Figure 3

   On the second run of the Dijkstra algorithm, Algorithm, the second path
   S->B->C->D with metric 6 is obtained.

   As mentioned in Section 8.5, the Multipath Dijkstra Algorithm does
   not guarantee strict disjoint paths in order to avoid choosing
   inferior paths.  For example, given the topology in Figure 4, two
   paths from node S to D are desired.  On the top of the figure, there
   is a high cost path between S and D.

   If an algorithm tries to obtain strict disjoint paths, the two paths
   obtained will be S--B--D and S--(high cost path)--D, which are
   extremely unbalanced.  It is undesirable because it will cause huge
   delay variance between the paths.  By using the Multipath Dijkstra
   algorithm,
   Algorithm, which is based on the punishing scheme, S--B--D and S--B--
   C--D will be obtained.

                             --high cost path-
                            /                 \
                           /                   \
                           S----B--------------D
                                 \           /
                                  \---C-----/

                                 Figure 4

Acknowledgments

   The authors would like to thank Sylvain David, Asmaa Adnane, Eddy
   Cizeron, Salima Hamma, Pascal Lesage, and Xavier Lecourtier for their
   efforts in developing, implementing, and testing the specification.
   The authors also appreciate valuable discussions with Thomas Clausen,
   Ulrich Herberg, Justin Dean, Geoff Ladwig, Henning Rogge, Marcus
   Barkowsky, and especially Christopher Dearlove for his multiple
   rounds of reviews during the working group last calls.

Authors' Addresses

   Jiazi Yi
   Ecole Polytechnique
   91128 Palaiseau Cedex
   France

   Phone: +33 (0) 1 77 57 80 85
   Email: jiazi@jiaziyi.com
   URI:   http://www.jiaziyi.com/

   Benoit Parrein
   University of Nantes
   IRCCyN Lab - IVC team
   Polytech Nantes, rue Christian Pauc, BP50609
   44306 Nantes cedex 3
   France

   Phone: +33 (0) 2 40 68 30 50
   Email: Benoit.Parrein@polytech.univ-nantes.fr
   URI:   http://www.irccyn.ec-nantes.fr/~parrein