idnits 2.17.1 draft-decraene-rtgwg-backoff-algo-01.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year == The document doesn't use any RFC 2119 keywords, yet seems to have RFC 2119 boilerplate text. -- The document date (March 9, 2015) is 3336 days in the past. Is this intentional? Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) No issues found here. Summary: 0 errors (**), 0 flaws (~~), 2 warnings (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group B. Decraene 3 Internet-Draft Orange 4 Intended status: Standards Track March 9, 2015 5 Expires: September 10, 2015 7 Back-off SPF algorithm for link state IGP 8 draft-decraene-rtgwg-backoff-algo-01 10 Abstract 12 This document defines a standard algorithm to back-off link-state IGP 13 SPF computations. 15 Having one standardized algorithm improves interoperability by 16 reducing the probability and/or duration of transient forwarding 17 loops during the IGP convergence in the area/level when the network 18 reacts to multiple consecutive events. 20 Requirements Language 22 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 23 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 24 document are to be interpreted as described in [RFC2119]. 26 Status of This Memo 28 This Internet-Draft is submitted in full conformance with the 29 provisions of BCP 78 and BCP 79. 31 Internet-Drafts are working documents of the Internet Engineering 32 Task Force (IETF). Note that other groups may also distribute 33 working documents as Internet-Drafts. The list of current Internet- 34 Drafts is at http://datatracker.ietf.org/drafts/current/. 36 Internet-Drafts are draft documents valid for a maximum of six months 37 and may be updated, replaced, or obsoleted by other documents at any 38 time. It is inappropriate to use Internet-Drafts as reference 39 material or to cite them other than as "work in progress." 41 This Internet-Draft will expire on September 10, 2015. 43 Copyright Notice 45 Copyright (c) 2015 IETF Trust and the persons identified as the 46 document authors. All rights reserved. 48 This document is subject to BCP 78 and the IETF Trust's Legal 49 Provisions Relating to IETF Documents 50 (http://trustee.ietf.org/license-info) in effect on the date of 51 publication of this document. Please review these documents 52 carefully, as they describe your rights and restrictions with respect 53 to this document. Code Components extracted from this document must 54 include Simplified BSD License text as described in Section 4.e of 55 the Trust Legal Provisions and are provided without warranty as 56 described in the Simplified BSD License. 58 1. Introduction 60 Link state IGP, such as IS-IS [ISO10589-Second-Edition] and OSPF 61 [RFC2328], performs distributed computation on all nodes of the area/ 62 level. In order to have consistent routing tables across the 63 network, such distributed computation requires that all routers have 64 the same vision of the network (Link State DataBase (LSDB)) and 65 perform their computation at the same time. 67 In general, when the network is stable, there is a desire to compute 68 the new SPF as soon as the failure is known, in order to quickly 69 route around the failure. However, when the network is experiencing 70 multiple consecutive failures over a short period of time, there is a 71 desire to limit the frequency of SPF computations. Indeed, this 72 allow reducing the control plane resources used by IGP and all 73 protocols/sub system reacting on it such as LDP, RSVP-TE, BGP, Fast 74 ReRoute computations, FIB updates..., reduce the churn on nodes and 75 in the network, in particular reduce side effects such as micro-loops 76 which may happen during each IGP convergence. 78 To allow for this, some back-off algorithm have been implemented. 79 Different implementations choose different algorithms, hence in a 80 multi-vendor network, it's not possible to enforce that all routers 81 triggers their SPF computation after the same waiting delay. This 82 situation increases the average differential delay between routers 83 end of RIB computation. It also increases the probability that 84 different routers compute their RIB based on a different LSDB. Both 85 increases the probability and/or duration of micro-loops. 87 To allow for multi-vendors networks having all the routers delaying 88 their SPF for the same duration, this document specifies a 89 standardized algorithm. Implementations may offer alternative 90 optional algorithms. 92 2. High level goals 94 The high level goals of this algorithm are the following: 96 o Very fast convergence for single simple events (link failure). 98 o Fast convergence in general while the IGP stability is considered 99 under control. 101 o A long delay when the IGP stability is considered out of control, 102 in order to let all related process calm down. 104 o At any time, try to avoid using different SPF_TIMERS values for 105 nodes in the area/level. Even though not all nodes will receive 106 IGP message at the same time (due to difference in distance from 107 the source and due to different flooding implementations on the 108 path from the source). 110 3. Definitions and parameters 112 IGP events: An LSDB change requiring a new RIB computation (topology 113 change, prefix change, metric change). No distinction is done 114 between the type of computation performed (e.g. full SPF, incremental 115 SPF, PRC). The type of computation is a local consideration. 117 The SPF_DELAY timer can take the following values: 119 INITIAL_WAIT: a very small delay to quickly handle link failure. 120 e.g. 0 millisecond. 122 FAST_WAIT: a small delay to have a fast convergence. e.g. 50-100 123 millisecond. Note: we want to be fast, but as this failure requires 124 multiple IGP events, being too fast increase the probability to 125 receive additional IGP events just after the RIB computation. 127 LONG_WAIT: a long delay as IGP is unstable. e.g. 2 seconds. Note: 128 let's bring calm in the IGP. 130 The TIME_TO_CONVERGE timer is the time to learn all the IGP events 131 related to a single failure (e.g. node failure, SRLG failure). e.g. 1 132 second. It's mostly dependent on variation of failure detection 133 times between all nodes which are neighbour to the failure, and then 134 may depend on different flooding algorithms of nodes in the network. 136 The HOLD_DOWN timer is the time needed with no IGP events received, 137 before considering that the IGP is quiet again and we can set the 138 SPF_DELAY back to INITAL_WAIT. e.g. 5 seconds. 140 4. Principle of SPF delay algorithm 142 The first IGP event is handled very quickly (INITIAL_WAIT) in order 143 to be very reactive for the first event if it only needs one IGP 144 event (e.g. link failure, prefix change). 146 If more IGP events are received quickly after, we consider that they 147 are related to the same single failure, and handle the IGP events 148 relatively quickly (FAST_WAIT) during the time needed to receive all 149 the IGP events related to the failure (TIME_TO_CONVERGE). 151 If IGP events are still received after this time, then the network is 152 presumably experiencing multiple independent failures and the while 153 waiting for its stability, the computations are delayed for a longer 154 time (LONG_WAIT). 156 Note: previous SPF delay algorithms used to count the number of RIB 157 computations. However, as all nodes may receive the LSP events in a 158 different way we cannot assume that all nodes will perform the same 159 number of SPF computations or that they will schedule them at the 160 same time. For example, assuming that the SPF delay is 50 ms, node 161 R1 may receive 3 IGP events (E1, E2, E3) in those 50 ms and hence 162 will perform a single routing computation. While another node R2 may 163 only receive 2 events (E1, E2) in those 50ms and hence will schedule 164 another routing computation when further receiving E3. That's why 165 this document prefers to define a time limit (TIME_TO_CONVERGE) since 166 the first event, rather than a number of routing computations. 168 5. Specification of SPF delay algorithm 170 When the previous IGP events is more than HOLD_DOWN ago: 172 o The IGP is set to the QUIET state. 174 When the IGP is in the QUIET state and an IGP event is received: 176 o The time of this first IGP event is stored in FIRST_EVENT_TIME. 178 o The next RIB computation time is set to LSP receive time + 179 INITIAL_WAIT. 181 o The IGP is set to the FAST_WAIT state. 183 When the IGP is in the FAST_WAIT state and an IGP event is received: 185 o If more than TIME_TO_CONVERGE has passed since FIRST_EVENT_TIME, 186 then the IGP is set to the HOLD_DOWN state. 188 o If the next RIB_computation time is in the past, set the next RIB 189 computation time to LSP receive time + FAST_WAIT. 191 When the IGP is in the HOLD_DOWN state and an IGP event is received: 193 o If the next RIB_computation time is in the past, set the next RIB 194 computation time to LSP receive time + LONG_WAIT. 196 6. Impact on micro-loops 198 Micro-loops during IGP convergence are due to a non synchronized or 199 non ordered update of the forwarding information tables (FIB) 200 [RFC5715] [RFC6976] [I-D.litkowski-rtgwg-spf-uloop-pb-statement]. 201 FIB are installed after multiple steps such as SPF wait time, SPF 202 computation, FIB distribution and FIB update. This document only 203 address the first contribution. This standardized procedure reduces 204 the probability and/or duration of micro-loops when the IGP 205 experience multiple consecutive events. It does not remove all 206 micro-loops. However, it is beneficial and its cost seems limited 207 compared to full solutions such as [RFC5715] or [RFC6976]. 209 7. IANA Considerations 211 No IANA actions required. 213 8. Security considerations 215 This document has no impact on the security of the IGP. 217 9. Acknowledgements 219 We would like to acknowledge Hannes Gredler, Les Ginsberg and Pierre 220 Francois for the discussions related to this document. 222 10. References 224 10.1. Normative References 226 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 227 Requirement Levels", BCP 14, RFC 2119, March 1997. 229 10.2. Informative References 231 [I-D.litkowski-rtgwg-spf-uloop-pb-statement] 232 Litkowski, S., "Link State protocols SPF trigger and delay 233 algorithm impact on IGP microloops", draft-litkowski- 234 rtgwg-spf-uloop-pb-statement-02 (work in progress), March 235 2015. 237 [ISO10589-Second-Edition] 238 International Organization for Standardization, 239 "Intermediate system to Intermediate system intra-domain 240 routeing information exchange protocol for use in 241 conjunction with the protocol for providing the 242 connectionless-mode Network Service (ISO 8473)", ISO/IEC 243 10589:2002, Second Edition, Nov 2002. 245 [RFC2328] Moy, J., "OSPF Version 2", STD 54, RFC 2328, April 1998. 247 [RFC5715] Shand, M. and S. Bryant, "A Framework for Loop-Free 248 Convergence", RFC 5715, January 2010. 250 [RFC6976] Shand, M., Bryant, S., Previdi, S., Filsfils, C., 251 Francois, P., and O. Bonaventure, "Framework for Loop-Free 252 Convergence Using the Ordered Forwarding Information Base 253 (oFIB) Approach", RFC 6976, July 2013. 255 Author's Address 257 Bruno Decraene 258 Orange 259 38 rue du General Leclerc 260 Issy Moulineaux cedex 9 92794 261 France 263 Email: bruno.decraene@orange.com