Internet Engineering Task Force (IETF)                        C. Pelsser
Request for Comments: 7196                                       R. Bush
Category: Standards Track                      Internet Initiative Japan
ISSN: 2070-1721                                                 K. Patel
                                                          Cisco Systems
                                                           P. Mohapatra
                                                       Sproute Networks
                                                             O. Maennel
                                                Loughborough University
                                                               May 2014


                   Making Route Flap Damping Usable

Abstract

  Route Flap Damping (RFD) was first proposed to reduce BGP churn in
  routers.  Unfortunately, RFD was found to severely penalize sites for
  being well connected because topological richness amplifies the
  number of update messages exchanged.  Many operators have turned RFD
  off.  Based on experimental measurement, this document recommends
  adjusting a few RFD algorithmic constants and limits in order to
  reduce the high risks with RFD.  The result is damping a non-trivial
  amount of long-term churn without penalizing well-behaved prefixes'
  normal convergence process.

Status of This Memo

  This is an Internet Standards Track document.

  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).  Further information on
  Internet Standards is available in Section 2 of RFC 5741.

  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/rfc7196.












Pelsser, et al.              Standards Track                    [Page 1]

RFC 7196            Making Route Flap Damping Usable            May 2014


Copyright Notice

  Copyright (c) 2014 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.  Suggested Reading . . . . . . . . . . . . . . . . . . . .   3
  2.  Requirements Language . . . . . . . . . . . . . . . . . . . .   3
  3.  RFD Parameters  . . . . . . . . . . . . . . . . . . . . . . .   3
  4.  Suppress Threshold versus Churn . . . . . . . . . . . . . . .   4
  5.  Maximum Penalty . . . . . . . . . . . . . . . . . . . . . . .   4
  6.  Recommendations . . . . . . . . . . . . . . . . . . . . . . .   5
  7.  Security Considerations . . . . . . . . . . . . . . . . . . .   5
  8.  Acknowledgments . . . . . . . . . . . . . . . . . . . . . . .   5
  9.  References  . . . . . . . . . . . . . . . . . . . . . . . . .   6
    9.1.  Normative References  . . . . . . . . . . . . . . . . . .   6
    9.2.  Informative References  . . . . . . . . . . . . . . . . .   6

1.  Introduction

  Route Flap Damping (RFD) was first proposed (see [RIPE178] and
  [RFC2439]) and subsequently implemented to reduce BGP churn in
  routers.  Unfortunately, RFD was found to severely penalize sites for
  being well connected because topological richness amplifies the
  number of update messages exchanged, see [MAO2002].  Subsequently,
  many operators turned RFD off; see [RIPE378].  Based on the
  measurements of [PELSSER2011], [RIPE580] now recommends that RFD is
  usable with some changes to the parameters.  Based on the same
  measurements, this document recommends adjusting a few RFD
  algorithmic constants and limits.  The result is damping of a non-
  trivial amount of long-term churn without penalizing well-behaved
  prefixes' normal convergence process.

  Very few prefixes are responsible for a large amount of the BGP
  messages received by a router; see [HUSTON2006] and [PELSSER2011].
  For example, the measurements in [PELSSER2011] showed that only 3% of



Pelsser, et al.              Standards Track                    [Page 2]

RFC 7196            Making Route Flap Damping Usable            May 2014


  the prefixes were responsible for 36% percent of the BGP messages at
  a router with real feeds from a Tier-1 provider and an Internet
  Exchange Point during a one-week experiment.  Only these very
  frequently flapping prefixes should be damped.  The values
  recommended in Section 6 achieve this.  Thus, RFD can be enabled, and
  some churn reduced.

  The goal is to, with absolutely minimal change, ameliorate the danger
  of current RFD implementations and use.  It is not a panacea, nor is
  it a deep and thorough approach to flap reduction.

1.1.  Suggested Reading

  It is assumed that the reader understands BGP [RFC4271] and Route
  Flap Damping [RFC2439].  This work is based on the measurements in
  the paper [PELSSER2011].  A survey of Japanese operators' use of RFD
  and their desires is reported in [RFD-SURVEY].

2.  Requirements Language

  The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
  "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" are to
  be interpreted as described in RFC 2119 [RFC2119] only when they
  appear in all upper case.  They may also appear in lower or mixed
  case as English words, without normative meaning.

3.  RFD Parameters

  The following RFD parameters are common to all implementations.  Some
  may be tuned by the operator, some not.  There is currently no
  consensus on a single set of default values.

        +--------------------------+----------+-------+---------+
        | Parameter                | Tunable? | Cisco | Juniper |
        +--------------------------+----------+-------+---------+
        | Withdrawal               | No       | 1,000 |   1,000 |
        | Re-Advertisement         | No       |     0 |   1,000 |
        | Attribute Change         | No       |   500 |     500 |
        | Suppress Threshold       | Yes      | 2,000 |   3,000 |
        | Half-Life (min.)         | Yes      |    15 |      15 |
        | Reuse Threshold          | Yes      |   750 |     750 |
        | Max Suppress Time (min.) | Yes      |    60 |      60 |
        +--------------------------+----------+-------+---------+

    Note: Values without units specified are dimensionless constants.

          Table 1: Default RFD Parameters of Juniper and Cisco




Pelsser, et al.              Standards Track                    [Page 3]

RFC 7196            Making Route Flap Damping Usable            May 2014


4.  Suppress Threshold versus Churn

  By turning RFD back on with the values recommended in Section 6,
  churn is reduced.  Moreover, with these values, prefixes going
  through normal convergence are generally not damped.

  [PELSSER2011] estimates that, with a suppress threshold of 6,000, the
  BGP update rate is reduced by 19% compared to a situation without RFD
  enabled.  [PELSSER2011] studies the number of prefixes damped over a
  week between September 29, 2010 and October 6, 2010.  With this 6,000
  suppress threshold, 90% fewer prefixes are damped compared to use of
  a 2,000 threshold.  That is, far fewer well-behaved prefixes are
  damped.

  Setting the suppress threshold to 12,000 leads to very few damped
  prefixes (0.22% of the prefixes were damped with a threshold of
  12,000 in the experiments in [PELSSER2011], yielding an average
  hourly update reduction of 11% compared to not using RFD).

  +---------------+-------------+--------------+----------------------+
  |      Suppress |      Damped |   % of Table |    Update Rate (one- |
  |     Threshold |    Prefixes |       Damped |           hour bins) |
  +---------------+-------------+--------------+----------------------+
  |         2,000 |      43,342 |       13.16% |               53.11% |
  |         4,000 |      11,253 |        3.42% |               74.16% |
  |         6,000 |       4,352 |        1.32% |               81.03% |
  |         8,000 |       2,104 |        0.64% |               84.85% |
  |        10,000 |       1,286 |        0.39% |               87.12% |
  |        12,000 |         720 |        0.22% |               88.74% |
  |        14,000 |         504 |        0.15% |               89.97% |
  |        16,000 |         353 |        0.11% |               91.01% |
  |        18,000 |         311 |        0.09% |               91.88% |
  |        20,000 |         261 |        0.08% |               92.69% |
  +---------------+-------------+--------------+----------------------+

     Note: the current default Suppress Threshold (2,000) is overly
                               agressive.

         Table 2: Damped Prefixes vs. Churn, from [PELSSER2011]

5.  Maximum Penalty

  It is important to understand that the parameters shown in Table 1
  and the implementation's sampling rate impose an upper bound on the
  penalty value, which we can call the 'computed maximum penalty'.






Pelsser, et al.              Standards Track                    [Page 4]

RFC 7196            Making Route Flap Damping Usable            May 2014


  In addition, BGP implementations have an internal constant, which we
  will call the 'maximum penalty', and the current computed penalty may
  not exceed it.

6.  Recommendations

  Use of the following values is recommended:

  Router Maximum Penalty:  The internal constant for the maximum
     penalty value MUST be raised to at least 50,000.

  Default Configurable Parameters:  In order not to break existing
     operational configurations, existing BGP implementations,
     including the examples in Table 1, SHOULD NOT change their default
     values.

  Minimum Suppress Threshold:  Operators that want damping that is much
     less destructive than the current damping, but still somewhat
     aggressive, SHOULD configure the Suppress Threshold to no less
     than 6,000.

  Conservative Suppress Threshold:  Conservative operators SHOULD
     configure the Suppress Threshold to no less than 12,000.

  Calculate But Do Not Damp:  Implementations MAY have a test mode
     where the operator can see the results of a particular
     configuration without actually damping any prefixes.  This will
     allow for fine-tuning of parameters without losing reachability.

7.  Security Considerations

  It is well known that an attacker can generate false flapping to
  cause a victim's prefix(es) to be damped.

  As the recommendations merely change parameters to more conservative
  values, there should be no increase in risk.  In fact, the parameter
  change to more conservative values should slightly mitigate the
  false-flap attack.

8.  Acknowledgments

  Nate Kushman initiated this work some years ago.  Ron Bonica, Seiichi
  Kawamura, and Erik Muller contributed useful suggestions.








Pelsser, et al.              Standards Track                    [Page 5]

RFC 7196            Making Route Flap Damping Usable            May 2014


9.  References

9.1.  Normative References

  [MAO2002]  Mao, Z., Govidan, R., Varghese, G., and R. Katz, "Route
             Flap Damping Exacerbates Internet Routing Convergence", In
             Proceedings of SIGCOMM, August 2002,
             <http://conferences.sigcomm.org/sigcomm/2002/papers/
             routedampening.pdf>.

  [PELSSER2011]
             Pelsser, C., Maennel, O., Mohapatra, P., Bush, R., and K.
             Patel, "Route Flap Damping Made Usable", PAM 2011: Passive
             and Active Measurement Conference, March 2011,
             <http://pam2011.gatech.edu/papers/pam2011--Pelsser.pdf>.

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

  [RFC2439]  Villamizar, C., Chandra, R., and R. Govindan, "BGP Route
             Flap Damping", RFC 2439, November 1998.

  [RFC4271]  Rekhter, Y., Li, T., and S. Hares, "A Border Gateway
             Protocol 4 (BGP-4)", RFC 4271, January 2006.

  [RIPE378]  Smith, P. and P. Panigl, "RIPE Routing Working Group
             Recommendations On Route-flap Damping", RIPE 378, May
             2006, <http://www.ripe.net/ripe/docs/ripe-378>.

9.2.  Informative References

  [HUSTON2006]
             Huston, G., "2005 - A BGP Year in Review", RIPE 52, 2006,
             <http://meetings.ripe.net/ripe-52/presentations/
             ripe52-plenary-bgp-review.pdf>.

  [RFD-SURVEY]
             Tsuchiya, S., Kawamura, S., Bush, R., and C. Pelsser,
             "Route Flap Damping Deployment Status Survey", Work in
             Progress, June 2012.

  [RIPE178]  Barber, T., Doran, S., Karrenberg, D., Panigl, C., and J.
             Schmitz, "RIPE Routing-WG Recommendation for Coordinated
             Route-flap Damping Parameters", RIPE 178, February 1998,
             <http://www.ripe.net/ripe/docs/ripe-178>.






Pelsser, et al.              Standards Track                    [Page 6]

RFC 7196            Making Route Flap Damping Usable            May 2014


  [RIPE580]  Bush, R., Pelsser, C., Kuhne, M., Maennel, O., Mohapatra,
             P., Patel, K., and R. Evans, "RIPE Routing Working Group
             Recommendation for Route Flap Damping", RIPE 580, January
             2013, <http://www.ripe.net/ripe/docs/ripe-580>.















































Pelsser, et al.              Standards Track                    [Page 7]

RFC 7196            Making Route Flap Damping Usable            May 2014


Authors' Addresses

  Cristel Pelsser
  Internet Initiative Japan
  Jinbocho Mitsui Buiding, 1-105
  Kanda-Jinbocho, Chiyoda-ku, Tokyo  101-0051
  JP

  Phone: +81 3 5205 6464
  EMail: [email protected]


  Randy Bush
  Internet Initiative Japan
  5147 Crystal Springs
  Bainbridge Island, Washington  98110
  US

  EMail: [email protected]


  Keyur Patel
  Cisco Systems
  170 W. Tasman Drive
  San Jose, CA  95134
  US

  EMail: [email protected]


  Pradosh Mohapatra
  Sproute Networks
  41529 Higgins Way
  Fremont, CA  94539
  US

  EMail: [email protected]


  Olaf Maennel
  Loughborough University
  Department of Computer Science - N.2.03
  Loughborough
  UK

  Phone: +44 115 714 0042
  EMail: [email protected]




Pelsser, et al.              Standards Track                    [Page 8]