[Docs] [txt|pdf] [Tracker] [Email] [Nits] [IPR]

Versions: 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 RFC 6971

Internet Engineering Task Force                              A. Cardenas
Internet-Draft                                      Fujitsu Laboratories
Intended status: Informational                               S. Cespedes
Expires: September 9, 2011                        University of Waterloo
                                                                 T. Iwao
                                                         Fujitsu Limited
                                                           March 8, 2011

             Depth-First Forwarding in Unreliable Networks


   Routing protocols are generally composed of two independent phases,
   the control plane and the data forwarding plane.  The control plane
   is responsible for route discovery and maintenance.  The data
   forwarding plane performs a table lookup operation to set the packet
   on the right path.  In unreliable networks, the routing process
   incurs a large control overhead when is constantly repairing routes,
   detecting loops, and finding alternate paths due to frequent link

   This document describes the Depth-First Forwarding (DFF) protocol; a
   data forwarding mechanism that can be used to minimize the burden and
   control overhead of a control plane used in unreliable networks.  DFF
   offers reliability and low control overhead by supporting in the data
   plane loop detection, updates to the routing tables, and rerouting of
   data packets through alternate paths.  DFF can be integrated with
   different types of control plane mechanisms and can be used in mesh-
   under and route-over specifications.  In this draft, we describe a
   sample DFF implementation as a 6LoWPAN mesh-under data forwarding

Status of this Memo

   This Internet-Draft is submitted in full conformance with the
   provisions of BCP 78 and BCP 79.

   Internet-Drafts are working documents of the Internet Engineering
   Task Force (IETF).  Note that other groups may also distribute
   working documents as Internet-Drafts.  The list of current Internet-
   Drafts is at http://datatracker.ietf.org/drafts/current/.

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

Cardenas, et al.        Expires September 9, 2011               [Page 1]

Internet-Draft                     DFF                        March 2011

   This Internet-Draft will expire on September 9, 2011.

Copyright Notice

   Copyright (c) 2011 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.

Cardenas, et al.        Expires September 9, 2011               [Page 2]

Internet-Draft                     DFF                        March 2011

Table of Contents

   1.  Introduction . . . . . . . . . . . . . . . . . . . . . . . . .  4
     1.1.  Requirements notation  . . . . . . . . . . . . . . . . . .  4
     1.2.  Terminology  . . . . . . . . . . . . . . . . . . . . . . .  4
   2.  Protocol Overview  . . . . . . . . . . . . . . . . . . . . . .  5
   3.  Hop-by-Hop Implementation Options  . . . . . . . . . . . . . .  5
   4.  Depth-First Forwarding Operation . . . . . . . . . . . . . . .  6
   5.  Message Formats  . . . . . . . . . . . . . . . . . . . . . . .  8
   6.  Data Structures  . . . . . . . . . . . . . . . . . . . . . . .  9
   7.  Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 10
   8.  IANA Considerations  . . . . . . . . . . . . . . . . . . . . . 10
   9.  Security Considerations  . . . . . . . . . . . . . . . . . . . 10
   10. Appendix A: Example Implementation of a Control Plane for
       DFF  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
   11. Appendix B: Implementing DFF without requesting new
       dispatch bytes . . . . . . . . . . . . . . . . . . . . . . . . 11
   12. Normative References . . . . . . . . . . . . . . . . . . . . . 12
   Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 12

Cardenas, et al.        Expires September 9, 2011               [Page 3]

Internet-Draft                     DFF                        March 2011

1.  Introduction

   Networks with dynamic links present a challenge for typical routing
   protocols because the reliability of links may be different at the
   time when the route was discovered, and at the time when data is

   In these unreliable networks, the control overhead for detecting
   routing errors and for fixing paths happens often, so it is important
   to avoid expensive control plane mechanisms that might overreact in
   the presense of instability.  Because a lightweight control plane
   mechanism cannot guarantee the construction and maintenance of error-
   free routes, a data forwarding protocol designed for these conditions
   should be able to detect errors and find backup paths to survive link

   This document describes Depth-First Forwarding (DFF), a data
   forwarding mechanism that can detect loops, update the routing
   tables, and reroute data packets via alternate paths.  DFF is
   compatible with light-weight control plane mechanisms supporting
   routing tables that maintain more than one possible next hop for each
   final destination.

1.1.  Requirements notation

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   document are to be interpreted as described in [RFC2119].

1.2.  Terminology

   Readers are expected to be familiar with all the terms and concepts
   that are discussed in "Transmission of IPv6 Packets over IEEE
   802.15.4 Networks" [RFC4944].

   Other terms used:

   Final Destination: This is the final destination of the data packet
   within the mesh network.

   Local destination: The local destination of the data packet refers to
   the next-hop neighbor to which the packet is forwarded on its way to
   the final destination.

   Originator: This is the source node that created the 6lowpan data

Cardenas, et al.        Expires September 9, 2011               [Page 4]

Internet-Draft                     DFF                        March 2011

2.  Protocol Overview

   DFF is a data forwarding strategy responsible for loop detection,
   choosing alternate next hops, and updating the cost metrics in the
   routing tables to reflect information gathered by forwarding data

   DFF is intended to work in a network where nodes maintain proactively
   a routing table with multiple candidate next hops for each final
   destination.  An example of a control plane satisfying these
   conditions is described in the Appendix.

   DFF provides an advantage in networks where the reliability of links
   changes rapidly.  It assumes that the control plane mechanism cannot
   guarantee up to date routing tables, nor the absence of loops.
   Therefore, whenever a data packet is forwarded, DFF can keep a data
   packet identifier to detect loops, update routing tables if a loop is
   detected, and use alternate paths to reroute the packet around the
   failed path.

   DFF achieves this functionality by implementing a distributed depth-
   first search over the network graph as defined by the routing table.
   If the routing tables are up to date, the search will only involve
   the default route.  However, if the routing table is not up to date
   and forwarding of a data packet results in a loop, or if the link
   layer fails to successfully transmit the packet to the next hop, the
   data packet is then sent to an alternate next hop neighbor.  A
   distributed depth-first search mechanism is implemented in order to
   keep track of the nodes that have participated in the forwarding of
   the data packet.

   Although DFF can be used without a control plane by performing a
   blind (i.e., without a routing table) depth-first search of the
   network, this configuration will incurr in increased latency because
   data packets are forwarded by intermediate nodes to a random next-hop
   neighbor.  Therefore, it is recommended to implement DFF in
   combination with a proactive control plane protocol, in order to
   efficiently guide the depth-first search by using information stored
   in the routing tables.

3.  Hop-by-Hop Implementation Options

   While DFF can be used in a route-over or mesh-under protocol, this
   document provides a sample implementation of a mesh-under forwarding
   solution for 6LoWPAN networks; therefore, all addresses referenced in
   this document are either 16-bit short or EUI-64 link layer addresses.

Cardenas, et al.        Expires September 9, 2011               [Page 5]

Internet-Draft                     DFF                        March 2011

   DFF requires the use of hop-by-hop options, and this document
   describes how these hop-by-hop options can be implemented by
   allocating a new dispatch byte from the reserved values for mesh
   forwarding in [RFC4944].

   To avoid the request of a new dispatch byte, the appendix describes a
   way to implement DFF by overloading the fragmentation header in
   [RFC4944].  This implementation has the advantage of using headers
   already defined by the standard; however, the implementation by
   overloading the fragmentation header only allows rerouting a packet
   on loop detection and not when a link fails.  The reason behind this
   loss of functionality is that rerouting when a link fails requires
   the use of a duplicate flag in the header of the packet.  This is a
   hop-by-hop option that can be dynamically updated by intermediate
   nodes, and the fragmentation header of [RFC4944] cannot be changed by
   intermediate nodes.

   Similarly, a route over implementation of DFF would need to obtain
   new fields in the hop-by-hop options of IPv6 packets.

4.  Depth-First Forwarding Operation

   The operation procedure described in this section relies on the
   existence of a routing table in every node.  This table SHOULD be
   filled by a proactive control plane that stores multiple candidate
   next hops for final destinations.  An example of a proactive distance
   vector control plane that could be integrated to DFF is provided in
   the Appendix.

   In order for an Originator to send packets based on depth-first
   forwarding, it encapsulates the data packet using the standard mesh
   header defined in [RFC4944] and the DFF mesh header (Figure 1).  The
   DFF mesh header is employed to detect loops and reroute packets in
   the forwarding path.  The Originator then checks the routing table to
   select the next hop with the lowest cost to reach the destination.
   Before forwarding the packet, an entry is created in the loop
   detection table (Figure 2), where information such as the
   Originator's address, data packet identifier, previous hop (it points
   to the node itself when the node is the Originator of packet), and
   the selected next hop are stored.

   Upon reception of a data packet at an intermediate node (which might
   be the Originator if there is a loop in the path), the node checks if
   an entry with the same (Originator,DID) exists in the loop detection
   table.  If there is no such entry, intermediate nodes SHOULD create a
   new entry in a similar way to that described for the originator of
   the data packet; however, in the Previous hop field, they store the

Cardenas, et al.        Expires September 9, 2011               [Page 6]

Internet-Draft                     DFF                        March 2011

   address of the router from which the packet was received.  After the
   entry has been created in the loop detection table, the node forwards
   the packet to the selected candidate next hop.

   For those cases in which an entry already exists in the loop
   detection table, the node checks which one was the last attemped
   node, and poisons the routing table entry that uses that particular
   node to reach the destination.  By poisoning failed paths, DFF
   updates the routing table based on the results from the data plane.
   Then, in order to reroute the data packet, the node selects a new
   next hop among the list of candidates stored in the routing table.
   The selected node MUST not be registered as a previous attempt in the
   list of attempted neighbours in the loop detection table.  I also
   MUST be a different node from that registered in the Previous Hop
   field.  In this way, DFF effectively makes data forwarding with loops
   a depth-first search guided by the routing table stored in each node.

   If the node has attempted all the candidate next hops, then the
   packet is returned to Previous Hop. If the Previous hop address is
   the same address of the current node, that means the node is the
   originator of the packet.  If in addition, the node has already
   attempted all next-hop options, this means that routing has failed;
   therefore, the originator must drop the packet and delete the entry
   in the loop table.

   In addition to rerouting packets when a loop is detected, nodes
   reroute packets when the link layer fails to receive ACK from the
   neighbor they sent the data packet to.  As soon as the link layer
   gives up on the transmission, DFF proceeds to reroute the packet
   through a different candidate.  In this case, nodes set a duplicate
   detection flag in the DFF mesh header, identifying whether or not the
   packet is a potential duplicate.  Duplicate packets can appear in the
   network when the link layer reports a failed transmission due to a
   failed reception of ACKs from the recipient of the packet.  This
   situation may appear on links that are lossy only in one direction.

   Duplicate packets do not alter the depth-first search logic: if a
   packet with a duplicate flag is received by a node who has already
   sent a packet with the same (Originator,DID) to Next Hop n (Last Next
   Hop attempted), it assumes that this resulted in a loop, and the node
   then attempts to reroute the packet to Next Hop n+1 (if available),
   or to send it back to Previous Hop if no other candidate next hop are
   available.  This, however, may be a false loop detection, therefore
   the node does not poison entries in the routing table whenever the
   forwarded packet has the duplicate flag activated.

   For packets encapsulated according to [RFC4944] that do not include a
   DFF mesh header, the DFF node processes them with a simple forwarding

Cardenas, et al.        Expires September 9, 2011               [Page 7]

Internet-Draft                     DFF                        March 2011

   mechanism that selects the next hop with the lowest cost to reach the
   final destination.  In this case, the node does node create any
   entries in the loop detection table, and it does not attempt to
   reroute such packets through alternate paths.  This forwarding option
   allows for the coexistence of DFF nodes with nodes that do not follow
   the message formats defined in this document (Figure 1).  A 6lowpan
   mesh header [RFC4944] is still required for the operation of this
   basic forwarding mechanism.

5.  Message Formats

   This document assumes that multi-hop forwarding occurs in the
   adaptation layer following the message format of [RFC4944].
   [RFC4944] indicates that hop-by-hop processing headers with
   additional mesh routing capabilities may be expressed by defining
   additional headers that precede fragmentation or addressing headers.
   Hence, all data packets to be forwarded using DFF MUST be preceded by
   the standard mesh (L2) addressing header defined in [RFC4944], and
   MAY be preceded by a header that identifies the data forwarding
   mechanism (in this case DFF).

   After these two headers, other LoWPAN headers such as hop-by-hop
   options, header compression or fragmentation can also be included
   before the actual payload.  (Figure 1) shows the mesh headers of a
   data frame to be forwarded with DFF.

                          1                   2                   3
      0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      |                   Mesh type and header
      |0 1|Mesh Forw|Flags|         DID             |

                 Figure 1: Mesh Header for DFF data frames

   Field definitions are as follows:

   Mesh type and header:  The mesh (L2) addressing header and its
      associated dispatch byte as defined in [RFC4944].

   Mesh Forw:  is a 6-bit identifier that allows for the use of
      different mesh forwarding mechanisms.  As specified in [RFC4944],
      additional mesh forwarding mechanisms should use the reserved
      dispatch byte values following LOWPAN_BCO; therefore, 01 SHOULD
      precede Mesh Forw.  A possible value to use as a mesh forwarding
      identifier based on the reserved ranges defined in [RFC4944] is

Cardenas, et al.        Expires September 9, 2011               [Page 8]

Internet-Draft                     DFF                        March 2011

      010001.  In this case the dispatch byte would be 01010001.

   Flags:  Bits in this field are used by DFF to set control flags. 0xx
      means that this current packet is not a duplicate, and 1xx is used
      to identify a potential duplicate.  The last two bits are reserved
      to define possible new options for data forwarding.

   DID:   This is Data Frame Identifier.  It is a sequence number
      generated by the Originator.  The originator address concatenated
      with the DID sequence number form an identifier of previously seen
      data packets.

6.  Data Structures

   The loop detection option is based on the idea of storing the DID and
   originator-ID of a data packet, so that if a packet containing the
   same DID identifier and originator is received, DFF detects it as a

   After the loop is detected, DFF follows a distributed depth-first
   search for the destination through the candidate next hops kept in
   the routing table.  In order to do a Depth-First search, nodes need
   to keep a list of their children (i.e., the candidate next hops that
   have been used to forward the packet), and the previous hop (the node
   who sent the data packet for the first time to the current router).

   A Loop Detection Table (Figure 2) needs to be kept by the nodes to
   support the loop detection functionality.  The candidate next hop
   field does not need to be pre-stored, it can be filled dynamically as
   soon as the node attempts to send the packet to a next-hop neighbor.

Cardenas, et al.        Expires September 9, 2011               [Page 9]

Internet-Draft                     DFF                        March 2011

    |Parameter| Description                                           |
    |(O,DID)  | Source Address concatenated with a sequential         |
    |         | number. Used to identify previously seen data packets |
    |Previous | Address of the router who sent the data packet for the|
    |Hop      | first time to the current router. If forwarding fails,|
    |         | return data packet to this router                     |
    |TTL      | Time to live for the current DID entry                |
    |Next Hop | First neighbor selected to forward the packet         |
    | 1       |                                                       |
    | ...     | ...                                                   |
    |Next hop | Neighbor selected the k-th time                       |
    | K       |                                                       |

            Figure 2: Basic Elements of a Loop Detection Table

7.  Acknowledgements

   Ganesh Venkatesh, and Geoff Mulligan provided useful discussions
   which helped shape this document.

8.  IANA Considerations

   This memo includes the request of a new dispatch byte to identify DFF
   headers.  In the Appendix there is an implementation that avoids the
   use of new dispatch bytes.

9.  Security Considerations

   The security of a mesh forwarding protocol depends on the integrity,
   authentication, and confidentiality of the messages.  The security
   mechanisms for protecting the network can be provided by link-layer
   technologies.  Further details are presented in the Security
   Considerations section of [RFC4944].

Cardenas, et al.        Expires September 9, 2011              [Page 10]

Internet-Draft                     DFF                        March 2011

10.  Appendix A: Example Implementation of a Control Plane for DFF

   There are many route discovery protocols compatible with DFF.  The
   final selection of which control plane to use depends on the
   particular constraints and requirements of the network.  For example,
   if nodes have tight memory constraints and the network is large,
   managing the size of the routing table is important.  Therefore, a
   control plane that builds a network with a routing table that grows
   at a slower rate than the size of the network--e.g., via hierarchical
   routing, or clustering--is important.  If minimizing the routing
   stretch of the network is a priority, then the control plane needs to
   keep larger routing tables.

   The only condition for a control plane in order to leverage an
   implementation of DFF is that, nodes should maintain a number of
   alternate routes, which are being advertised by multiple neighbors
   and which can be used immediately if the selected route were to fail,
   or if a loop is detected through a previous route.

   While a number of routing protocols satisfy the above constraint,
   they tend to include extra overhead for preventing loops or dealing
   with routing inconsistencies or failures.  One of the primary goals
   of DFF is to avoid the use of these extra control messages.  This
   appendix presents a basic control plane compatible with DFF.


11.  Appendix B: Implementing DFF without requesting new dispatch bytes

   DFF can be implemented as a full standard conforming to [RFC4944]
   without requesting any new dispatch bytes.  In this way, nodes
   implementing DFF can interoperate with other nodes that only
   implement headers defined in [RFC4944].

   A possible way to avoid the DFF mesh header is by overloading the
   datagram_tag and datagram_offset fields of the fragmentation header
   defined in [RFC4944].

   Because each source maintains a sequence number for the datagram_tag,
   and the datagram_offset can be used to differentiate between
   fragmented packets with the same value in datagram_tag, the DID value
   required by DFF can be generated by the concatenation of the
   datagram_tag and datagram_offset values of a fragmented data frame.

   Nonetheless, an implementation of DFF that avoids the request of a
   new dispatch byte will prevent the use of flags, and without the
   existence of a duplicate flag, duplicate packets will not be

Cardenas, et al.        Expires September 9, 2011              [Page 11]

Internet-Draft                     DFF                        March 2011

   detected.  Therefore, it is RECOMMENDED that nodes that implement DFF
   by using the datagram_tag and datagram_offset fields for storing the
   DID value, do not reroute on link-layer ACK failures, but only on
   loop detections.  In this case, all previously seen (Originator,DID)
   values can be assumed to correspond to loop detections, and the
   routing table cost to reach the final destination via the last
   attempted neighbor can be safely poisoned, without the risk of
   poisoning valid routes taken by duplicate packets.

   This implementation of DFF assumes the existence of fragmentation
   headers within the LoWPAN encapsulation.  This works well if data
   packets are fragmented, but if the entire payload datagram fits
   within a single 802.15.4 frame, then [RFC4944] states that the LoWPAN
   encapsulation should not contain a fragmentation header.  However,
   the use of a fragmentation header for a packet that does not need to
   be fragmented should, in principle, not affect the operation of nodes
   implementing [RFC4944].  Therefore, even if a packet does not need to
   be fragmented, the originator node can append the fragmentation
   header so DFF nodes can use it for extracting the DID identifier.

   The control plane used to populate the routing tables can also avoid
   the need to request a new dispatch byte by encapsulating routing
   updates in UDP packets.

12.  Normative References

   [RFC2119]  Bradner, S., "Key words for use in RFCs to Indicate
              Requirement Levels", BCP 14, RFC 2119, March 1997.

   [RFC4944]  Montenegro, G., Kushalnagar, N., Hui, J., and D. Culler,
              "Transmission of IPv6 Packets over IEEE 802.15.4
              Networks", RFC 4944, September 2007.

Authors' Addresses

   Alvaro A. Cardenas
   Fujitsu Laboratories
   1240 E. Arques Avenue, M/S 345
   Sunnyvale, CA  94085

   Phone: +1 408 530-4516
   Email: alvaro.cardenas-mora@us.fujitsu.com

Cardenas, et al.        Expires September 9, 2011              [Page 12]

Internet-Draft                     DFF                        March 2011

   Sandra Cespedes
   University of Waterloo
   200 University Ave. W.
   Waterloo, ON  N2L 3G1

   Phone: +1 (519) 8884567 x37448
   Email: slcesped@bbcr.uwaterloo.ca

   Tadashige Iwao
   Fujitsu Limited
   Fujitsu Kyushu R and D Center, 2-1, Momochihama 2-chome, Sawara-ku.

   Phone: +81-92-821-8030
   Email: smartnetpro-iwao_std@ml.css.fujitsu.com

Cardenas, et al.        Expires September 9, 2011              [Page 13]

Html markup produced by rfcmarkup 1.107, available from http://tools.ietf.org/tools/rfcmarkup/