roll                                                     P. van der Stok
Internet-Draft                                                consultant
Intended status: Standards Track                               AR. Sangi
Expires: June 10, 2017                               Huawei Technologies
                                                        December 7, 2016


                      MPL Forwarder Select (MPLFS)
                   draft-ietf-roll-mpl-forw-select-00

Abstract

   This document describes a Forwarder Selection (MPLFS) protocol for
   the Multicast Protocol for Low-Power and lossy Networks (MPL) to
   reduce the density of forwarders such that the number of forwarded
   messages is reduced.

   The protocol uses Trickle to distribute link-local information about
   the identity of the neighbours of the nodes that have MPL-enabled
   interfaces.  In the end-state all nodes are connected to a minimum
   number, N_DUPLICATE, of forwarders, where N_DUPLICATE is application
   dependent, and there is a path between any two forwarders.

Note

   Discussion and suggestions for improvement are requested, and should
   be sent to roll@ietf.org.

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."

   This Internet-Draft will expire on June 10, 2017.







van der Stok & Sangi      Expires June 10, 2017                 [Page 1]


Internet-Draft                    MPLFS                    December 2016


Copyright Notice

   Copyright (c) 2016 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  . . . . . . . . . . . . . . . . . . . . . . . .   2
     1.1.  Terminology . . . . . . . . . . . . . . . . . . . . . . .   3
   2.  Protocol overview . . . . . . . . . . . . . . . . . . . . . .   4
   3.  Data sets . . . . . . . . . . . . . . . . . . . . . . . . . .   4
   4.  Neighbor distribution . . . . . . . . . . . . . . . . . . . .   5
   5.  Selection Algorithm . . . . . . . . . . . . . . . . . . . . .   6
   6.  CBOR payload  . . . . . . . . . . . . . . . . . . . . . . . .   7
   7.  Default parameter values  . . . . . . . . . . . . . . . . . .   7
   8.  Acknowledgements  . . . . . . . . . . . . . . . . . . . . . .   8
   9.  Changelog . . . . . . . . . . . . . . . . . . . . . . . . . .   8
   10. References  . . . . . . . . . . . . . . . . . . . . . . . . .   8
     10.1.  Normative References . . . . . . . . . . . . . . . . . .   8
     10.2.  Informative References . . . . . . . . . . . . . . . . .   9
   Appendix A.  example forwarder allocations  . . . . . . . . . . .   9
   Authors' Addresses  . . . . . . . . . . . . . . . . . . . . . . .  10

1.  Introduction

   The Multicast Protocol for Low-Power and Lossy Networks (MPL)
   [RFC7731] is designed for small devices interconnected by a lossy
   wireless network such as IEEE 802.15.4.  A seed sends a multicast
   message with a realm-local scope, admin-local scope or higher as
   specified in [RFC4291].

   Forwarders forward these messages with an increasing interval size.
   When the density of the forwarders is high, the DATA_MESSAGE_K (k)
   parameter stops the retransmission of messages in a given Trickle
   interval when more than k copies of a message have been received.
   This mechanism prevents network saturation but leads to difficult to
   predict end-to-end data transmission delays.  For IoT short
   predictable delays are wanted and the k parameter is set to a large



van der Stok & Sangi      Expires June 10, 2017                 [Page 2]


Internet-Draft                    MPLFS                    December 2016


   value.  When the density of forwarders is high and k is large, the
   message may be forwarded by a high number of forwarders that conflict
   on the link.  With extreme forwarder densities, small Trickle
   intervals, and k is infinite, just sending one multicast message may
   lead to an overload of the communication medium.

   The number of forwarded messages can be reduced by selecting a
   minimal set of forwarders.  However, for large networks, manually
   selecting the forwarders is much work, and changing network
   conditions and configurations make the manual selection an unwanted
   burden to the network management.

   This document specifies a protocol that selects the forwarders such
   that each MPL-enabled device is connected to N_DUPLICATE forwarders,
   where N_DUPLICATE > 0 can be set.  The parameter N_DUPLICATE
   determines how much path redundancy there is for each MPL message.
   The value of N_DUPLICATE should be at least 1, because a value of 0
   has as result that no forwarder exists in the network during the
   protocol execution.  Moreover, the protocol is distributed and
   dynamic in nature to face a continuously changing topology.

   The protocol is inspired by the work described for NeighbourHood
   Discovery (NHDP) [RFC6130] and Simple Multicast Forwarding (SMF)
   [RFC6621].  In contrast to the "HELLO" messages described in
   [RFC6130], this protocol uses the Trickle protocol [RFC6206] to
   multicast link-local messages, containing a CBOR payload [RFC7049].

   A simulation of the protocol has been done for some example
   topologies.  The number of forwarders allocated by the protocol are
   compared in Appendix A with the optimum (lowest) number of forwarders
   for these topologies.

1.1.  Terminology

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
   document are to be interpreted as described in [RFC2119].

   Readers of this specification should be familiar with all the terms
   and concepts discussed in [RFC7731].  The following terms are defined
   in this document:

   synchronization time  The moment that a node can change its state at
      messages reception.

   The following list contains the abbreviations used in this document.

   XXXX:  TODO, and others to follow.



van der Stok & Sangi      Expires June 10, 2017                 [Page 3]


Internet-Draft                    MPLFS                    December 2016


2.  Protocol overview

   Nodes participating in MPLFS exchange messages with a format that is
   described in Section 6.  A participating node communicates to all its
   neighbours with link-local multicast messages as described in
   Section 4.

   Failing links provide a lot of instability.  Only messages sent over
   stable links are accepted.  Section 4 describes a mechanism to refuse
   messages from unstable links.

   Each node maintains a set of 1-hop neighbours where each neighbour
   contains information about its own 1-hop neighbours.  On the basis of
   the contents of the set, the node can decide to become a Forwarder or
   not, as explained in Section 5.

   The protocol never ends, with a minimum frequency of exchanging
   maintenance messages specified by an interval size of I_MAX_SELECT.
   When the set of links is stable, the protocol stabilizes such that
   there is a path between any two forwarders, and every MPL-enabled
   node is connected to at least N_DUPLICATE MPL forwarders (when
   existing), where N_DUPLICATE > 0.  N_DUPLICATE can be set dependent
   on the application requirements.  With N_DUPLICATE = 2, it is
   expected that a multicast message arrives at an intended recipient
   with very high probability.

   Nodes have a state that determines whether they are forwarder or not.
   The state of a node can only be changed by the node itself.  To avoid
   race conditions, (e.g. two nodes simultaneously decide to be
   forwarder, while only one is intended) the node with the highest
   address of all 1-hop neighbours is the only one allowed to change
   state.  Unlike [RFC5614], that considers 3-tuple (Router Priority,
   MDR Level and Router ID) to allow self state change, this approach
   only takes into account the node address.  Consequently, only k-hop
   neighbours, with k > 2, can change state simultaneously, and the
   1-hop and 2-hop neighbours of a given node can change state one by
   one.

3.  Data sets

   Each node, n_0, maintains a state with two values: Fixed Forwarder
   (FF) and No Forwarder (NF).  Each node also maintains a set, S1_0,
   containing information about n_0's 1-hop neighbours and n_0 itself.
   Each entry, n_i, in S1_0 has the following attributes:

   address of n_i:   the address can be the 64 bit IPv6 address or the
      short 16 bit address.




van der Stok & Sangi      Expires June 10, 2017                 [Page 4]


Internet-Draft                    MPLFS                    December 2016


   average-rssi-in:   the average rssi of the messages received by n_0
      from n_i.

   average-rssi-out:   the average rssi of the messages received by n_i
      from n_0.

   nr_FF:   the number of neighbours, n_ij, of n_i (including n_i) with
      state = FF.

   nr_Under:   the number of neighbours, n_ij, of n_i with nr_FF <
      N_DUPLICATE.

   nr_Above:   the number of neighbours, n_ij, of n_i with nr_FF >
      N_DUPLICATE.

   size:   the size of S1_i, the set of 1-hop neighbours of n_i.

   state:   the state of n_i.

4.  Neighbor distribution

   A participating node multicasts link-local so-called "neighbour
   messages" with the Trickle protocol.  It uses the multicast address
   LINK_LOCAL_ALL_NODES as destination.  The message sent by n_0
   contains the contents of S1_0.  The contents of a "neighbour message"
   from n_i received by n_0 is called M_i.  The rssi value associated
   with the reception of the "neighbour message" is called new_rssi.
   The message M_i contains information about the set S1_i with the
   following attributes for all nodes in S1_i:

   o  address

   o  average-rssi-in

   o  nr_FF

   o  nr_Under

   o  nr_Above

   o  size

   o  state

   On reception of M_i from n_i for the first time, the receiving node
   adds n_i to S1_0, and sets average-rssi-in of n_i in S1_0 to
   new_rssi.  For all following messages from n_i, the average-rssi-in




van der Stok & Sangi      Expires June 10, 2017                 [Page 5]


Internet-Draft                    MPLFS                    December 2016


   for n_i is calculated in the following way: average-rssi-in :=
   (average-rssi-in*WEIGHT_AVERAGE + new_rssi)/(WEIGHT_AVERAGE+1).

   The neighbour nodes of M_i are called n_ij.  For the n_ij with an
   address that is equal to the address of n_0: the value of average-
   rssi-out of n_0 is set equal to the value of average-rssi-in of n_ij.

   The contents of n_0 is updated with the contents of M_i.  Updating
   includes the following actions:

   o  Add n_i to S1_0, if n_i not present in S1_0.

   o  Set size of n_i equal to the number of entries in M_i.

   o  When n_ij.address = n_j.address, copy the values of nr_Under,
      nr_Above, nr_FF, and state of n_ij to n_j.

   When the average-rssi-in and average-rssi-out values of n_i have been
   averaged over more than WEIGHT_AVERAGE messages, and the averaged
   RSSI values are smaller than MAXIMUM_RSSI, n_i is called "valid".

5.  Selection Algorithm

   The protocol aims at allocating forwarders in the densest part of the
   network.  A dense network is characterized by a high number of
   neighbours.  Therefore, the protocol attempts to assign status FF to
   the nodes with the highest number of neighbours that have less than
   N_DUPLICATE neighbours with state = FF (nr_FF < N_DUPLICATE).

   It is required that a path exists between every two forwarders to
   prevent network partitioning.  Therefore, a node can become forwarder
   iff one of its neighbours is a forwarder.  The consequence of this
   rule is that one so-called "source-forwarder" must be selected by the
   network administrator.  A likely choice for the "source-forwarder" is
   the border router.

   At the start of the selection protocol the node, n_0, sets its state
   to No Forwarder (NF).  It sets the Trickle timer to its minimum
   interval, I_MIN_SELECT, and starts multicasting M_0 to its
   neighbours.  Every time entries are added to, or removed from, S1_0,
   the Trickle interval timer is set to I_MIN_SELECT.

   The executing node, n_0, calculates the following parameter values:

   o  max-under is the maximum of the nr_Under attribute of all valid
      n_i in S1_0.





van der Stok & Sangi      Expires June 10, 2017                 [Page 6]


Internet-Draft                    MPLFS                    December 2016


   o  max_address_u is the maximum of the addresses of valid n_i with
      nr_Under = max-under.

   o  max_address_a is the maximum of the addresses of all valid n_i.

   o  connected is true iff nr_FF of all neighbouring forwarders is
      equal to nr_FF of n_0.

   The information about the state and the nr_Under value of the
   neighbours comes in asynchronously.  Time is needed before the state
   in a node correctly reflects the state changes of the network.  A
   node can change its state when during the reception of messages of
   all neighbours, the value of nr_Under has not changed.

   To calculate its new state, n_0 does the following:

   When the state is NF, a neighbour with state = FF exists, and address
    = max-address_u:
      set state to FF.

   When the state is FF, nr_Above = size S1_0, connected is true, and
   address = max_address_a:
      set state to NF.

6.  CBOR payload

   The payload format is /application/cbor [RFC7049].  The contents of
   the message is a CBOR array (Major type 4) of CBOR arrays composed of
   neighbour address, rssi value, size of S1_i, forwarder state, nr_FF,
   nr_Under, and nr_Above.  Assuming two neighbours, in diagnostic JSON
   the payload looks like:

   [
   [address_1, average-rssi-in_1, size_1, state_1,
       nr_FF_1, nr_Under_1, nr_Above_1],
   [address_2, average-rssi-in_2, size_2, state_2,
       nr_FF_2, nr_Under_2, nr_Above_2]
   ]

                          Figure 1: CBOR payload

7.  Default parameter values

   The following text recommends default values for the MPLFS protocols.

   I_MIN_SELECT  = 0,2; minimum Trickle timer interval.

   I_MAX_SELECT  = 10; maximum Trickle timer interval.



van der Stok & Sangi      Expires June 10, 2017                 [Page 7]


Internet-Draft                    MPLFS                    December 2016


   DATA_MESSAGE_K  > 10; the redundancy constant.

   WEIGHT_AVERAGE  = 10; number of values to average rssi.

   MAXIMUM_RSSI  = 3; maximum acceptable average rssi value.

   N_DUPLICATE  = 2; requested number of MPL forwarder neighbours for
      every MPL enabled node.

8.  Acknowledgements

   We are very grateful to Yasuyuki Tanaka for pointing out the missing
   discussion on k-values.

9.  Changelog

   Changes from vanderstok version to ietf version 00

   o  copied from vanderstok-roll-forw-select-01

   o  added discussion on DATA_MESSAGE_K

   o  compared generated forwarder set with optmal set

10.  References

10.1.  Normative References

   [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>.

   [RFC6206]  Levis, P., Clausen, T., Hui, J., Gnawali, O., and J. Ko,
              "The Trickle Algorithm", RFC 6206, DOI 10.17487/RFC6206,
              March 2011, <http://www.rfc-editor.org/info/rfc6206>.

   [RFC7049]  Bormann, C. and P. Hoffman, "Concise Binary Object
              Representation (CBOR)", RFC 7049, DOI 10.17487/RFC7049,
              October 2013, <http://www.rfc-editor.org/info/rfc7049>.

   [RFC7731]  Hui, J. and R. Kelsey, "Multicast Protocol for Low-Power
              and Lossy Networks (MPL)", RFC 7731, DOI 10.17487/RFC7731,
              February 2016, <http://www.rfc-editor.org/info/rfc7731>.







van der Stok & Sangi      Expires June 10, 2017                 [Page 8]


Internet-Draft                    MPLFS                    December 2016


10.2.  Informative References

   [RFC4291]  Hinden, R. and S. Deering, "IP Version 6 Addressing
              Architecture", RFC 4291, DOI 10.17487/RFC4291, February
              2006, <http://www.rfc-editor.org/info/rfc4291>.

   [RFC5614]  Ogier, R. and P. Spagnolo, "Mobile Ad Hoc Network (MANET)
              Extension of OSPF Using Connected Dominating Set (CDS)
              Flooding", RFC 5614, DOI 10.17487/RFC5614, August 2009,
              <http://www.rfc-editor.org/info/rfc5614>.

   [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>.

   [RFC6621]  Macker, J., Ed., "Simplified Multicast Forwarding",
              RFC 6621, DOI 10.17487/RFC6621, May 2012,
              <http://www.rfc-editor.org/info/rfc6621>.

Appendix A.  example forwarder allocations

   The protocol has been simulated with omnet++ on four different
   network topologies.  The value of N_DUPLICATE is set to two.  Each
   topology is a rectangular grid, where each grid point is occupied by
   a node.  The distance, D, between two horizontal and vertical direct
   neighbours is the same.  The topologies differ in number of grid
   points and D.  A topology of 9x9 grid points shows how the selection
   criteria of the protocol affect the distribution of the forwarders
   when the number of edge nodes is small compared to the number of
   interior nodes.  A topology of 3x20 nodes shows the distribution with
   a majority of edge nodes.  Two radio ranges of the link have been
   selected: (1) a range of 3,5 times D, and (2) a range of 7 times D.
   The node density of case 2 is four times higher than for case 1.  The
   two example ranges nicely show how the average distance between the
   forwarders increases with increasing range.

   Table 1 shows the comparison results.  Column one indicates the grid
   topology, Columns 2 indicates the radio range, column 3 indicates the
   number of protocol forwarders as delivered by the simulation, column
   four indicates the minimum number of forwarders needed to arrive at
   at least N-DUPLICATE forwarders for all nodes, column five cites the
   number of nodes.








van der Stok & Sangi      Expires June 10, 2017                 [Page 9]


Internet-Draft                    MPLFS                    December 2016


           +----------+--------+------------+---------+-------+
           | topology | range  | forwarders | optimum | nodes |
           +----------+--------+------------+---------+-------+
           | 9x9      | 3,5 D  | 10         | 9       | 81    |
           |          |        |            |         |       |
           | 9x9      | 7 D    | 3          | 3       | 81    |
           |          |        |            |         |       |
           | 3x20     | 3,5 D  | 8          | 8       | 60    |
           |          |        |            |         |       |
           | 3x20     | 7 D    | 5          | 5       | 60    |
           +----------+--------+------------+---------+-------+

         Table 1: comparison of protocol set size and optimal set

   The table gives confidence that for many cases the size of the
   protocol forwarder set will be close to the size of an optimal set.

Authors' Addresses

   Peter van der Stok
   consultant

   Phone: +31-492474673 (Netherlands), +33-966015248 (France)
   Email: consultancy@vanderstok.org
   URI:   www.vanderstok.org


   Abdur Rashid Sangi
   Huawei Technologies
   No.156 Beiqing Rd. Haidian District
   Beijing  100095
   P.R. China

   Email: rashid.sangi@huawei.com

















van der Stok & Sangi      Expires June 10, 2017                [Page 10]