idnits 2.17.1 draft-ietf-tewg-diff-te-mar-03.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- ** Looks like you're using RFC 2026 boilerplate. This must be updated to follow RFC 3978/3979, as updated by RFC 4748. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- == The page length should not exceed 58 lines per page, but there was 13 longer pages, the longest (page 11) being 64 lines Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- ** There are 11 instances of too long lines in the document, the longest one being 4 characters in excess of 72. ** There are 36 instances of lines with control characters in the document. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the RFC 3978 Section 5.4 Copyright Line does not match the current year == Line 936 has weird spacing: '...cedures for...' -- The document seems to lack a disclaimer for pre-RFC5378 work, but may have content which was first submitted before 10 November 2008. If you have contacted all the original authors and they are all willing to grant the BCP78 rights to the IETF Trust, then this is fine, and you can ignore this comment. If not, you may need to add the pre-RFC5378 disclaimer. (See the Legal Provisions document at https://trustee.ietf.org/license-info for more information.) -- Couldn't find a document date in the document -- date freshness check skipped. Checking references for intended status: Experimental ---------------------------------------------------------------------------- == Missing Reference: 'RFC2119' is mentioned on line 69, but not defined == Missing Reference: 'IANA-CONS' is mentioned on line 435, but not defined == Unused Reference: 'KEY' is defined on line 450, but no explicit reference was found in the text == Unused Reference: 'ASH2' is defined on line 459, but no explicit reference was found in the text == Unused Reference: 'ASH3' is defined on line 461, but no explicit reference was found in the text == Unused Reference: 'RFC2026' is defined on line 497, but no explicit reference was found in the text Summary: 3 errors (**), 0 flaws (~~), 9 warnings (==), 2 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 1 Network Working Group Jerry Ash 2 Internet Draft AT&T 3 Category: Experimental 4 5 Expiration Date: July 2004 6 January, 2004 8 Max Allocation with Reservation Bandwidth Constraints Model for 9 DiffServ-aware MPLS Traffic Engineering & Performance Comparisons 11 13 Status of this Memo 15 This document is an Internet-Draft and is in full conformance with 16 all provisions of Section 10 of RFC 2026. 18 Internet-Drafts are working documents of the Internet Engineering 19 Task Force (IETF), its areas, and its working groups. Note that other 20 groups may also distribute working documents as Internet-Drafts. 22 Internet-Drafts are draft documents valid for a maximum of six months 23 and may be updated, replaced, or obsoleted by other documents at any 24 time. It is inappropriate to use Internet-Drafts as reference 25 material or to cite them other than as "work in progress." 27 The list of current Internet-Drafts can be accessed at 28 http://www.ietf.org/ietf/1id-abstracts.txt 30 The list of Internet-Draft Shadow Directories can be accessed at 31 http://www.ietf.org/shadow.html. 33 Abstract 35 This document complements the DiffServ-aware MPLS TE (DS-TE) requirements 36 document by giving a functional specification for the Maximum Allocation 37 with Reservation (MAR) Bandwidth Constraints Model. Assumptions, 38 applicability, and examples of the operation of the MAR Bandwidth 39 Constraints Model are presented. MAR performance is analyzed relative to 40 the criteria for selecting a Bandwidth Constraints Model, in order to 41 provide guidance to user implementation of the model in their networks. 43 Internet Draft MAR Bandwidth Constraints Model for DS-TE Jan 04 45 Table of Contents 47 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3 48 2. Definitions. . . . . . . . . . . . . . . . . . . . . . . . . . 4 49 3. Assumptions & Applicability . . . . . . . . . . . . . . . . . 5 50 4. Functional Specification of the MAR Bandwidth Constraints Model 5 51 5. Setting Bandwidth Constraints . . . . . . . . . . . . . . . . 6 52 6. Example of MAR Operation . . . . . . . . . . . . . . . . . . . 7 53 7. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 54 8. Security Considerations . . . . . . . . . . . . . . . . . . . 8 55 9. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 8 56 10. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 8 57 11. Normative References . . . . . . . . . . . . . . . . . . . . 9 58 12. Informative References . . . . . . . . . . . . . . . . . . . 9 59 13. Intellectual Property Statement . . . . . . . . . . . . . . . 10 60 14. Authors' Addresses . . . . . . . . . . . . . . . . . . . . . 10 61 Appendix A. MAR Operation & Performance Analysis . . . . . . . . 10 62 Appendix B. Bandwidth Prediction for Path Computation . . . . . . 16 64 Specification of Requirements 66 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 67 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 68 document are to be interpreted as described in [RFC2119]. 70 Internet Draft MAR Bandwidth Constraints Model for DS-TE Jan 04 72 1. Introduction 74 DiffServ-aware MPLS traffic engineering (DS-TE) requirements and protocol 75 extensions are specified in [DSTE-REQ, DSTE-PROTO]. A requirement for 76 DS-TE implementation is the specification of Bandwidth Constraints Models 77 for use with DS-TE. The Bandwidth Constraints Model provides the 'rules' 78 to support the allocation of bandwidth to individual class types (CTs). 79 CTs are groupings of service classes in the DS-TE model, which are 80 provided separate bandwidth allocations, priorities, and QoS objectives. 81 Several CTs can share a common bandwidth pool on an integrated, 82 multiservice MPLS/DiffServ network. 84 This document is intended to complement the DS-TE requirements document 85 [DSTE-REQ] by giving a functional specification for the Maximum 86 Allocation with Reservation (MAR) Bandwidth Constraints Model. Examples 87 of the operation of the MAR Bandwidth Constraints Model are presented. 88 MAR performance is analyzed relative to the criteria for selecting a 89 Bandwidth Constraints Model, in order to provide guidance to user 90 implementation of the model in their networks. 92 Two other Bandwidth Constraints Models are being specified for use in 93 DS-TE: 95 1. Maximum Allocation Model (MAM) [MAM] - the maximum allowable 96 bandwidth usage of each CT is explicitly specified. 97 2. Russian Doll Model (RDM) [RDM] - the maximum allowable bandwidth 98 usage is done cumulatively by grouping successive CTs according to 99 priority classes. 101 MAR is similar to MAM in that a maximum bandwidth allocation is given to 102 each CT. However, through the use of bandwidth reservation and 103 protection mechanisms, CTs are allowed to exceed their bandwidth 104 allocations under conditions of no congestion but revert to their 105 allocated bandwidths when overload and congestion occurs. 107 All Bandwidth Constraints Models should meet these objectives: 109 1. applies equally when preemption is either enabled or disabled (when 110 preemption is disabled, the model still works 'reasonably' well), 111 2. bandwidth efficiency, i.e., good bandwidth sharing among CTs under 112 both normal and overload conditions, 113 3. bandwidth isolation, i.e., a CT cannot hog the bandwidth of another 114 CT under overload conditions, 115 4. protection against QoS degradation, at least of the high-priority CTs 116 (e.g. high-priority voice, high-priority data, etc.), and 117 5. reasonably simple, i.e., does not require additional IGP extensions 118 and minimizes signaling load processing requirements. 120 In Appendix A modeling analysis is presented which shows that the MAR 121 Model meets all these objectives, and provides good network performance 122 relative to MAM and full sharing models, under normal and abnormal 123 operating conditions. It is demonstrated that simultaneously achieves 124 bandwidth efficiency, bandwidth isolation, and protection against QoS 125 degradation without preemption. 127 In Section 3 we give the assumptions and applicability, in Section 4 a 128 functional specification of the MAR Bandwidth Constraints Model, and in 129 Section 5 we give examples of its operation. In Appendix A, MAR 130 performance is analyzed relative to the criteria for selecting a 131 Internet Draft MAR Bandwidth Constraints Model for DS-TE Jan 04 133 Bandwidth Constraints Model, in order to provide guidance to user 134 implementation of the model in their networks. In Appendix B, 135 bandwidth prediction for path computation is discussed. 137 2. Definitions 139 For readability a number of definitions from [DSTE-REQ, DSTE-PROTO] are 140 repeated here: 142 Traffic Trunk: an aggregation of traffic flows of the same class (i.e. 143 which are to be treated equivalently from the DS-TE perspective) which 144 are placed inside an LSP. 146 Class-Type (CT): the set of Traffic Trunks crossing a link that is 147 governed by a specific set of Bandwidth constraints. CT is used for the 148 purposes of link bandwidth allocation, constraint based routing and 149 admission control. A given Traffic Trunk belongs to the same CT on all 150 links. 152 Up to 8 CTs (MaxCT = 8) are supported. They are referred to as CTc, 0 153 <= c <= MaxCT-1 = 7. Each CT is assigned either a Bandwidth 154 Constraint, or a set of Bandwidth Constraints. Up to 8 Bandwidth 155 Constraints (MaxBC = 8) are supported and they are referred to as BCc, 156 0 <= c <= MaxBC-1 = 7. 158 TE-Class: A pair of: i. a CT ii. a preemption priority allowed for that 159 CT. This means that an LSP transporting a Traffic Trunk from that CT can 160 use that preemption priority as the set-up priority, as the holding 161 priority or both. 163 MAX_RESERVABLE_BWk: maximum reservable bandwidth on link k specifies the 164 maximum bandwidth that may be reserved; this may be greater than the 165 maximum link bandwidth in which case the link may be oversubscribed 166 [OSPF-TE]. 168 BCck: bandwidth constraint for CTc on link k = allocated (minimum 169 guaranteed) bandwidth for CTc on link k (see Section 4). 171 RBW_THRESk: reservation bandwidth threshold for link k (see Section 4). 173 RESERVED_BWck: reserved bandwidth-in-progress on CTc on link k (0 <= c 174 <= MaxCT-1), RESERVED_BWck = total amount of the bandwidth reserved 175 by all the established LSPs which belong to CTc. 177 UNRESERVED_BWk: unreserved link bandwidth on link k specifies the 178 amount of bandwidth not yet reserved for any CT, UNRESERVED_BWk = 179 MAX_RESERVABLE_BWk - sum [RESERVED_BWck (0 <= c <= MaxCT-1)]. 181 UNRESERVED_BWck: unreserved link bandwidth on CTc on link k specifies 182 the amount of bandwidth not yet reserved for CTc, UNRESERVED_BWck = 183 MAX_RESERVABLE_BWk - UNRESERVED_BWk - delta0/1(CTck) * RBW-THRESk 184 where 185 delta0/1(CTck) = 0 if RESERVED_BWck < BCck 186 delta0/1(CTck) = 1 if RESERVED_BWck >= BCck 188 A number of recovery mechanisms under investigation in the IETF take 189 advantage of the concept of bandwidth sharing across particular sets of 190 LSPs. "Shared Mesh Restoration" in [GMPLS-RECOV] and "Facility-based 191 Computation Model" in [MPLS-BACKUP] are example mechanisms which 192 Internet Draft MAR Bandwidth Constraints Model for DS-TE Jan 04 194 increase bandwidth efficiency by sharing bandwidth across backup LSPs 195 protecting against independent failures. To ensure that the notion of 196 RESERVED_BWck introduced in [DSTE-REQ] is compatible with such a concept 197 of bandwidth sharing across multiple LSPs, the wording of the definition 198 provided in [DSTE-REQ] is generalized. With this generalization, the 199 definition is compatible with Shared Mesh Restoration defined in 200 [GMPLS-RECOV], so that DS-TE and Shared Mesh Protection can operate 201 simultaneously, under the assumption that Shared Mesh Restoration 202 operates independently within each DS-TE Class-Type and does not operate 203 across Class-Types. For example, backup LSPs protecting primary LSPs of 204 CTc must also belong to CTc; excess traffic LSPs sharing bandwidth with 205 backup LSPs of CTc must also belong to CTc. 207 3. Assumptions & Applicability 209 In general, DS-TE is a bandwidth allocation mechanism, for different 210 classes of traffic allocated to various CTs (e.g., voice, normal data, 211 best-effort data). Network operations functions such as capacity 212 design, bandwidth allocation, routing design, and network planning are 213 normally based on traffic measured load and forecast [ASH1]. 215 As such, the following assumptions are made according to the operation 216 of MAR: 218 1. connection admission control (CAC) allocates bandwidth for network 219 flows/LSPs according to the traffic load assigned to each CT, based on 220 traffic measurement and forecast. 221 2. CAC could allocate bandwidth per flow, per LSP, per traffic trunk, or 222 otherwise. That is, no specific assumption is made on a specific CAC 223 method, only that CT bandwidth allocation is related to the 224 measured/forecast traffic load, as per assumption #1. 225 3. CT bandwidth allocation is adjusted up or down according to 226 measured/forecast traffic load. No specific time period is assumed for 227 this adjustment, it could be short term (hours), daily, weekly, monthly, 228 or otherwise. 229 4. Capacity management and CT bandwidth allocation thresholds (e.g., 230 BCc) are designed according to traffic load, and are based on traffic 231 measurement and forecast. Again, no specific time period is assumed for 232 this adjustment, it could be short term (hours), daily, weekly, monthly, 233 or otherwise. 234 5. No assumption is made on the order in which traffic is allocated to 235 various CTs, again traffic allocation is assumed to be based only on 236 traffic load as it is measured and/or forecast. 237 6. If link bandwidth is exhausted on a given path for a flow/LSP/traffic 238 trunk, alternate paths may be attempted to satisfy CT bandwidth 239 allocation. 241 Note that the above assumptions are not unique to MAR, but are generic, 242 common assumptions for all BC Models. 244 4. Functional Specification of the MAR Bandwidth Constraints Model 246 A DS-TE LSR implementing MAR MUST support enforcement of bandwidth 247 constraints in compliance with the specifications in this Section. 249 In the MAR Bandwidth Constraints Model, the bandwidth allocation control 250 for each CT is based on estimated bandwidth needs, bandwidth use, and 251 status of links. The LER makes needed bandwidth allocation changes, and 252 uses [RSVP-TE], for example, to determine if link bandwidth can be 253 Internet Draft MAR Bandwidth Constraints Model for DS-TE Jan 04 255 allocated to a CT. Bandwidth allocated to individual CTs is protected as 256 needed but otherwise shared. Under normal non-congested network 257 conditions, all CTs/services fully share all available bandwidth. When 258 congestion occurs for a particular CTc, bandwidth reservation acts to 259 prohibit traffic from other CTs from seizing the allocated capacity for 260 CTc. 262 On a given link k, a small amount of bandwidth RBW_THRESk, the 263 reservation bandwidth threshold for link k, is reserved and governs the 264 admission control on link k. Also associated with each CTc on link k 265 are the allocated bandwidth constraints BCck to govern bandwidth 266 allocation and protection. The reservation bandwidth on a link, 267 RBW_THRESk, can be accessed when a given CTc has bandwidth-in-use 268 RESERVED_BWck below its allocated bandwidth constraint BCck. However, 269 if RESERVED_BWck exceeds its allocated bandwidth constraint BCck, then 270 the reservation bandwidth RBW_THRESk cannot be accessed. In this way, 271 bandwidth can be fully shared among CTs if available, but is otherwise 272 protected by bandwidth reservation methods. 274 Bandwidth can be accessed for a bandwidth request = DBW for CTc on a 275 given link k based on the following rules: 277 Table 1: Rules for Admitting LSP Bandwidth Request = DBW on Link k 279 For LSP on a high priority or normal priority CTc: 280 If RESERVED_BWck <= BCc: admit if DBW <= UNRESERVED_BWk 281 If RESERVED_BWck > BCc: admit if DBW <= UNRESERVED_BWk - RBW_THRESk; 282 or, equivalently: 283 If DBW <= UNRESERVED_BWck, admit the LSP. 285 For LSP on a best-effort priority CTc: 286 allocated bandwidth BCc = 0; 287 DiffServ queuing admits BE packets only if there is available link 288 bandwidth. 290 The normal semantics of setup and holding priority are applied in the 291 MAR Bandwidth Constraints Model, and cross-CT preemption is permitted 292 when preemption is enabled. 294 The bandwidth allocation rules defined in Table 1 are illustrated with 295 an example in Section 6 and simulation analysis in Appendix A. 297 5. Setting Bandwidth Constraints 299 For a normal priority CTc, the bandwidth constraints BCck on link k are 300 set by allocating the maximum reservable bandwidth (MAX_RESERVABLE_BWk) 301 in proportion to the forecast or measured traffic load bandwidth 302 TRAF_LOAD_BWck for CTc on link k. That is: 304 PROPORTIONAL_BWck = TRAF_LOAD_BWck/[sum {TRAF_LOAD_BWck, c=0,MaxCT-1}] X 305 MAX_RESERVABLE_BWk 307 For normal priority CTc: 308 BCck = PROPORTIONAL_BWck 309 Internet Draft MAR Bandwidth Constraints Model for DS-TE Jan 04 311 For a high priority CT, the bandwidth constraint BCck is set to a 312 multiple of the proportional bandwidth. That is: 314 For high priority CTc: 315 BCck = FACTOR X PROPORTIONAL_BWck 317 where FACTOR is set to a multiple of the proportional bandwidth (e.g., 318 FACTOR = 2 or 3 is typical). This results in some 'over-allocation' 319 of the maximum reservable bandwidth, and gives priority to the high 320 priority CTs. Normally the bandwidth allocated to high priority CTs 321 should be a relatively small fraction of the total link bandwidth, a 322 maximum of 10-15 percent being a reasonable guideline. 324 As stated in Section 4, the bandwidth allocated to a best-effort 325 priority CTc should be set to zero. That is: 327 For best-effort priority CTc: 328 BCck = 0 330 6. Example of MAR Operation 332 In the example, assume there are three class-types: CT0, CT1, CT2. We 333 consider a particular link with 335 MAX-RESERVABLE_BW = 100 337 And with the allocated bandwidth constraints set as follows: 339 BC0 = 30 340 BC1 = 20 341 BC2 = 20 343 These bandwidth constraints are based on the normal traffic loads, as 344 discussed in Section 5. With MAR, any of the CTs is allowed to exceed 345 its bandwidth constraint BCc as long a there is at least RBW_THRES 346 (reservation bandwidth threshold on the link) units of spare bandwidth 347 remaining. Let's assume 349 RBW_THRES = 10 351 So under overload, if 353 RESERVED_BW0 = 50 354 RESERVED_BW1 = 30 355 RESERVED_BW2 = 10 357 Therefore, for this loading 359 UNRESERVED_BW = 100 - 50 - 30 - 10 = 10 361 CT0 and CT1 can no longer increase their bandwidth on the link, since 362 they are above their BC values and there is only RBW_THRES=10 units of 363 spare bandwidth left on the link. But CT2 can take the additional 364 bandwidth (up to 10 units) if the demand arrives, since it is below its 365 BC value. 367 As also discussed in Section 4, if best effort traffic is present, it 368 can always seize whatever spare bandwidth is available on the link at 369 the moment, but is subject to being lost at the queues in favor of the 370 Internet Draft MAR Bandwidth Constraints Model for DS-TE Jan 04 372 higher priority traffic. 374 Let's say an LSP arrives for CT0 needing 5 units of bandwidth (i.e., DBW 375 = 5). We need to decide based on Table 1 whether to admit this LSP or 376 not. Since for CT0 378 RESERVED_BW0 > BC0 (50 > 30), and 379 DBW > UNRESERVED_BW - RBW_THRES (i.e., 5 > 10 - 10) 381 Table 1 says the LSP is rejected/blocked. 383 Now let's say an LSP arrives for CT2 needing 5 units of bandwidth (i.e., 384 DBW = 5). We need to decide based on Table 1 whether to admit this 385 LSP or not. Since for CT2 387 RESERVED_BW2 < BC2 (10 < 20), and 388 DBW < UNRESERVED_BW (i.e., 5 < 10) 390 Table 1 says to admit the LSP. 392 Hence, in the above example, in the current state of the link and the 393 current CT loading, CT0 and CT1 can no longer increase their bandwidth 394 on the link, since they are above their BCc values and there is only 395 RBW_THRES=10 units of spare bandwidth left on the link. But CT2 can 396 take the additional bandwidth (up to 10 units) if the demand arrives, 397 since it is below its BCc value. 399 7. Summary 401 The proposed MAR Bandwidth Constraints Model includes the following: a) 402 allocate bandwidth to individual CTs, b) protect allocated bandwidth by 403 bandwidth reservation methods, as needed, but otherwise fully share 404 bandwidth, c) differentiate high-priority, normal-priority, and 405 best-effort priority services, and d) provide admission control to 406 reject connection requests when needed to meet performance objectives. 407 Modeling results presented in Appendix A show that MAR bandwidth 408 allocation a) achieves greater efficiency in bandwidth sharing while 409 still providing bandwidth isolation and protection against QoS 410 degradation, and b) achieves service differentiation for high-priority, 411 normal-priority, and best-effort priority services. 413 8. Security Considerations 415 Security considerations related to the use of DS-TE are discussed in 416 [DSTE-PROTO]. Those apply independently of the Bandwidth Constraints 417 Model, including MAR specified in this document. 419 9. Acknowledgements 421 DS-TE and Bandwidth Constraints Models have been an active area of 422 discussion in the TEWG. I would like to thank Wai Sum Lai for his 423 support and review of this draft. I also appreciate helpful discussions 424 with Francois Le Faucheur. 426 10. IANA Considerations 428 [DSTE-PROTO] defines a new name space for "Bandwidth Constraints Model 429 Id". The guidelines for allocation of values in that name space are 430 detailed in Section 14 of [DSTE-PROTO]. In accordance with these 431 Internet Draft MAR Bandwidth Constraints Model for DS-TE Jan 04 433 guidelines, IANA was requested to assign a Bandwidth Constraints Model 434 Id for MAR from the range 0-127 (which is to be managed as per the 435 "Specification Required" policy defined in [IANA-CONS]). 437 Bandwidth Constraints Model Id = TBD was allocated by IANA to MAR. 439 To be removed by the RFC editor at the time of publication 440 We request IANA to assign value 2 for the MAR model. Once the value 441 has been assigned, please replace "TBD" above by the assigned value. 442 444 11. Normative References 446 [DSTE-REQ] Le Faucheur, F., Lai, W., et. al., "Requirements for Support 447 of Diff-Serv-aware MPLS Traffic Engineering," RFC 3564, July 2003. 448 [DSTE-PROTO] Le Faucheur, F., et. al., "Protocol Extensions for Support 449 of Diff-Serv-aware MPLS Traffic Engineering," work in progress. 450 [KEY] Bradner, S., "Key words for Use in RFCs to Indicate Requirement 451 Levels", RFC 2119, March 1997. 453 12. Informative References 455 [AKI] Akinpelu, J. M., "The Overload Performance of Engineered Networks 456 with Nonhierarchical & Hierarchical Routing," BSTJ, Vol. 63, 1984. 457 [ASH1] Ash, G. R., "Dynamic Routing in Telecommunications Networks," 458 McGraw-Hill, 1998. 459 [ASH2] Ash, G. R., et. al., "Routing Evolution in Multiservice Integrated 460 Voice/Data Networks," Proceeding of ITC-16, Edinburgh, June 1999. 461 [ASH3] Ash, G. R., "Performance Evaluation of QoS-Routing Methods for 462 IP-Based Multiservice Networks," Computer Communications Magazine, 463 May 2003. 464 TDM-Based Multiservice Networks, work in progress. 465 [BUR] Burke, P. J., Blocking Probabilities Associated with Directional 466 Reservation, unpublished memorandum, 1961. 467 [DSTE-PERF] Lai, W., "Bandwidth Constraints Models for DiffServ-TE: 468 Performance Evaluation", work in progress. 469 [E.360.1 --> E.360.7] ITU-T Recommendations, "QoS Routing & Related 470 Traffic Engineering Methods for Multiservice TDM-, ATM-, & IP-Based 471 Networks". 472 [GMPLS-RECOV] Lang, J., et. al., "Generalized MPLS Recovery Functional 473 Specification", work in progress. 474 [KRU] Krupp, R. S., "Stabilization of Alternate Routing Networks", 475 Proceedings of ICC, Philadelphia, 1982. 476 [LAI] Lai, W., "Traffic Engineering for MPLS, Internet Performance and 477 Control of Network Systems III Conference", SPIE Proceedings Vol. 4865, 478 pp. 256-267, Boston, Massachusetts, USA, 29 July-1 August 2002 479 (http://www.columbia.edu/~ffl5/waisum/bcmodel.pdf). 480 [MAM] Le Faucheur, F., Lai, W., "Maximum Allocation Bandwidth 481 Constraints Model for Diff-Serv-aware MPLS Traffic Engineering", work in 482 progress. 483 [MPLS-BACKUP] Vasseur, J. P., et. al., "MPLS Traffic Engineering Fast 484 Reroute: Bypass Tunnel Path Computation for Bandwidth Protection", work 485 in progress. 486 [MUM] Mummert, V. S., "Network Management and Its Implementation on the 487 No. 4ESS, International Switching Symposium", Japan, 1976. 488 [NAK] Nakagome, Y., Mori, H., Flexible Routing in the Global 489 Communication Network, Proceedings of ITC-7, Stockholm, 1973. 490 [OSPF-TE] Katz, D., et. al., "Traffic Engineering (TE) Extensions to 491 OSPF Version 2," RFC 3630, September 2003. 493 Internet Draft MAR Bandwidth Constraints Model for DS-TE Jan 04 495 [RDM] Le Faucheur, F., "Russian Dolls Bandwidth Constraints Model for 496 Diff-Serv-aware MPLS Traffic Engineering", work in progress. 497 [RFC2026] Bradner, S., "The Internet Standards Process -- Revision 3", 498 BCP 9, RFC 2026, October 1996. 499 [RSVP-TE] Awduche, D., et. al., "RSVP-TE: Extensions to RSVP for LSP 500 Tunnels", RFC 3209, December 2001. 502 13. Intellectual Property Statement 504 AT&T Corporation may own intellectual property applicable to this 505 contribution. The IETF has been notified of AT&T's licensing intent 506 for the specification contained in this document. See 507 http://www.ietf.org/ietf/IPR/ATT-GENERAL.txt for AT&T's IPR statement. 509 14. Authors' Addresses 511 Jerry Ash 512 AT&T 513 Room MT D5-2A01 514 200 Laurel Avenue 515 Middletown, NJ 07748, USA 516 Phone: +1 732-420-4578 517 Email: gash@att.com 519 Appendix A. MAR Operation & Performance Analysis 521 A.1 MAR Operation 523 In the MAR Bandwidth Constraints Model, the bandwidth allocation control 524 for each CT is based on estimated bandwidth needs, bandwidth use, and 525 status of links. The LER makes needed bandwidth allocation changes, and 526 uses [RSVP-TE], for example, to determine if link bandwidth can be 527 allocated to a CT. Bandwidth allocated to individual CTs is protected as 528 needed but otherwise shared. Under normal non-congested network 529 conditions, all CTs/services fully share all available bandwidth. When 530 congestion occurs for a particular CTc, bandwidth reservation acts to 531 prohibit traffic from other CTs from seizing the allocated capacity for 532 CTc. Associated with each CT is the allocated bandwidth constraint 533 (BCc) to govern bandwidth allocation and protection, these parameters 534 are illustrated with examples in this Appendix. 536 In performing MAR bandwidth allocation for a given flow/LSP, the LER 537 first determines the egress LSR address, service-identity, and CT. The 538 connection request is allocated an equivalent bandwidth to be routed on 539 a particular CT. The LER then accesses the CT priority, QoS/traffic 540 parameters, and routing table between the LER and egress LSR, and sets 541 up the connection request using the MAR bandwidth allocation rules. The 542 LER selects a first choice path and determines if bandwidth can be 543 allocated on the path based on the MAR bandwidth allocation rules given 544 in Section 4. If the first choice path has insufficient bandwidth, the 545 LER may then try alternate paths, and again applies the MAR bandwidth 546 allocation rules now described. 548 MAR bandwidth allocation is done on a per-CT basis, in which aggregated 549 CT bandwidth is managed to meet the overall bandwidth requirements of CT 550 service needs. Individual flows/LSPs are allocated bandwidth in the 551 corresponding CT according to CT bandwidth availability. A fundamental 552 principle applied in MAR bandwidth allocation methods is the use of 553 bandwidth reservation techniques. 555 Internet Draft MAR Bandwidth Constraints Model for DS-TE Jan 04 557 Bandwidth reservation gives preference to the preferred traffic by 558 allowing it to seize any idle bandwidth on a link, while allowing the 559 non-preferred traffic to only seize bandwidth if there is a minimum 560 level of idle bandwidth available called the reservation bandwidth 561 threshold RBW_THRES. Burke [BUR] first analyzed bandwidth reservation 562 behavior from the solution of the birth-death equations for the 563 bandwidth reservation model. Burke's model showed the relative 564 lost-traffic level for preferred traffic, which is not subject to 565 bandwidth reservation restrictions, as compared to non-preferred 566 traffic, which is subject to the restrictions. Bandwidth reservation 567 protection is robust to traffic variations and provides significant 568 dynamic protection of particular streams of traffic. It is widely used 569 in large-scale network applications [ASH1, MUM, AKI, KRU, NAK]. 571 Bandwidth reservation is used in MAR bandwidth allocation to control 572 sharing of link bandwidth across different CTs. On a given link, a 573 small amount of bandwidth RBW_THRES is reserved (say 1% of the total 574 link bandwidth), and the reservation bandwidth can be accessed when a 575 given CT has reserved bandwidth-in-progress RESERVED_BW below its 576 allocated bandwidth BC. That is, if the available link bandwidth 577 (unreserved idle link bandwidth UNRESERVED_BW) exceeds RBW_THRES, then 578 any CT is free to access the available bandwidth on the link. However, 579 if UNRESERVED_BW is less than RBW_THRES, then the CT can utilize the 580 available bandwidth only if its current bandwidth usage is below the 581 allocated amount BC. In this way, bandwidth can be fully shared among 582 CTs if available, but is protected by bandwidth reservation if below the 583 reservation level. 585 Through the bandwidth reservation mechanism, MAR bandwidth allocation 586 also gives preference to high-priority CTs, in comparison to 587 normal-priority and best-effort priority CTs. 589 Hence, bandwidth allocated to each CT is protected by bandwidth 590 reservation methods, as needed, but otherwise shared. Each LER monitors 591 CT bandwidth use on each CT, and determines if connection requests can 592 be allocated to the CT bandwidth. For example, for a bandwidth request 593 of DBW on a given flow/LSP, the LER determines the CT priority (high, 594 normal, or best-effort), CT bandwidth-in-use, and CT bandwidth 595 allocation thresholds, and uses these parameters to determine the 596 allowed load state threshold to which capacity can be allocated. In 597 allocating bandwidth DBW to a CT on given LSP, say A-B-E, each link in 598 the path is checked for available bandwidth in comparison to the allowed 599 load state. If bandwidth is unavailable on any link in path A-B-E, 600 another LSP could by tried, such as A-C-D-E. Hence determination of the 601 link load state is necessary for MAR bandwidth allocation, and two link 602 load states are distinguished: available (non-reserved) bandwidth 603 (ABW_STATE), and reserved-bandwidth (RBW_STATE). Management of CT 604 capacity uses the link state and the allowed load state threshold to 605 determine if a bandwidth allocation request can be accepted on a given 606 CT. 608 A.2 Analysis of MAR Performance 610 In this Appendix, modeling analysis is presented in which MAR bandwidth 611 allocation is shown to provide good network performance relative to full 612 sharing models, under normal and abnormal operating conditions. A 613 large-scale DiffServ-aware MPLS traffic engineering simulation model is 614 used, in which several CTs with different priority classes share the pool 615 of bandwidth on a multiservice, integrated voice/data network. MAR 616 Internet Draft MAR Bandwidth Constraints Model for DS-TE Jan 04 618 methods have also been analyzed in practice for TDM-based networks [ASH1], 619 and in modeling studies for IP-based networks [ASH2, ASH3, E.360]. 621 All Bandwidth Constraints Models should meet these objectives: 623 1. applies equally when preemption is either enabled or disabled (when 624 preemption is disabled, the model still works 'reasonably' well), 625 2. bandwidth efficiency, i.e., good bandwidth sharing among CTs under 626 both normal and overload conditions, 627 3. bandwidth isolation, i.e., a CT cannot hog the bandwidth of another 628 CT under overload conditions, 629 4. protection against QoS degradation, at least of the high-priority CTs 630 (e.g. high-priority voice, high-priority data, etc.), and 631 5. reasonably simple, i.e., does not require additional IGP extensions 632 and minimizes signaling load processing requirements. 634 The use of any given Bandwidth Constraints Model has significant impacts 635 on the performance of a network, as explained later. Therefore, the 636 criteria used to select a model must enable us to evaluate how a 637 particular model delivers its performance, relative to other models. Lai 638 [LAI, DSTE-PERF] has analyzed the MAM and RDM Models and provided 639 valuable insights into the relative performance of these models under 640 various network conditions. 642 In environments where preemption is not used, MAM is attractive because 643 a) it is good at achieving isolation, and b) it achieves reasonable 644 bandwidth efficiency with some QoS degradation of lower classes. When 645 preemption is used, RDM is attractive because it can achieve bandwidth 646 efficiency under normal load. However, RDM cannot provide service 647 isolation under high load or when preemption is not used. 649 Our performance analysis of MAR bandwidth allocation methods is based on 650 a full-scale, 135-node simulation model of a national network together 651 with a multiservice traffic demand model to study various scenarios and 652 tradeoffs [ASH3, E.360]. Three levels of traffic priority - high, 653 normal, and best effort -- are given across 5 CTs: normal priority voice, 654 high priority voice, normal priority data, high priority data, and best 655 effort data. 657 The performance analyses for overloads and failures include a) the MAR 658 Bandwidth Constraints Model, as specified in Section 4, b) the MAM 659 Bandwidth Constraints Model, and c) the No-DSTE Bandwidth Constraints 660 Model. 662 The allocated bandwidth constraints for MAR are as described in Section 663 5: 665 Normal priority CTs: BCck = PROPORTIONAL_BWk, 666 High priority CTs: BCck = FACTOR X PROPORTIONAL_BWk 667 Best-effort priority CTs: BCck = 0 669 In the MAM Bandwidth Constraints Model, the bandwidth constraints for 670 each CT are set to a multiple of the proportional bandwidth allocation: 672 Normal priority CTs: BCck = FACTOR1 X PROPORTIONAL_BWk, 673 High priority CTs: BCck = FACTOR2 X PROPORTIONAL_BWk 674 Best-effort priority CTs: BCck = 0 675 Internet Draft MAR Bandwidth Constraints Model for DS-TE Jan 04 677 Simulations show that for MAM, the sum (BCc) should exceed 678 MAX_RESERVABLE_BWk for better efficiency, as follows: 680 1. The normal priority CTs the BCc values need to be over-allocated to 681 get reasonable performance. It was found that over-allocating by 100%, 682 that is, setting FACTOR1 = 2, gave reasonable performance. 683 2. The high priority CTs can be over-allocated by a larger multiple 684 FACTOR2 in MAM and this gives better performance. 686 The rather large amount of over-allocation improves efficiency but 687 somewhat defeats the 'bandwidth protection/isolation' needed with a BC 688 Model, since one CT can now invade the bandwidth allocated to another 689 CT. Each CT is restricted to its allocated bandwidth constraint BCck, 690 which is the maximum level of bandwidth allocated to each CT on each 691 link, as in normal operation of MAM. 693 In the No-DSTE Bandwidth Constraints Model, no reservation or protection 694 of CT bandwidth is applied, and bandwidth allocation requests are 695 admitted if bandwidth is available. Furthermore, no queuing priority 696 is applied to any of the CTs in the No-DSTE Bandwidth Constraints Model. 698 Table 2 gives performance results for a six-times overload on a single 699 network node at Oakbrook IL. The numbers given in the table are the 700 total network percent lost (blocked) or delayed traffic. Note that in 701 the focused overload scenario studied here, the percent lost/delayed 702 traffic on the Oakbrook node is much higher than the network-wide 703 average values given. 705 Table 2 706 Performance Comparison for MAR, MAM, & No-DSTE 707 Bandwidth Constraints (BC) Models 708 6X Focused Overload on Oakbrook (Total Network % Lost/Delayed Traffic) 710 Class Type MAR BC MAM BC No-DSTE BC 711 Model Model Model 712 NORMAL PRIORITY VOICE 0.00 1.97 10.30 713 HIGH PRIORITY VOICE 0.00 0.00 7.05 714 NORMAL PRIORITY DATA 0.00 6.63 13.30 715 HIGH PRIORITY DATA 0.00 0.00 7.05 716 BEST EFFORT PRIORITY DATA 12.33 11.92 9.65 718 Clearly the performance is better with MAR bandwidth allocation, and the 719 results show that performance improves when bandwidth reservation is 720 used. The reason for the poor performance of the No-DSTE Model, without 721 bandwidth reservation, is due to the lack of protection of allocated 722 bandwidth. If we add the bandwidth reservation mechanism, then 723 performance of the network is greatly improved. 725 The simulations showed that the performance of MAM is quite sensitive to 726 the over-allocation factors discussed above. For example, if the BCc 727 values are proportionally allocated with FACTOR1 = 1, then the results 728 are much worse, as shown in Table 3: 730 Internet Draft MAR Bandwidth Constraints Model for DS-TE Jan 04 732 Table 3 733 Performance Comparison for MAM Bandwidth Constraints Model 734 with Different Over-allocation Factors 735 6X Focused Overload on Oakbrook (Total Network % Lost/Delayed Traffic) 737 Class Type (FACTOR1 = 1) (FACTOR1 = 2) 738 NORMAL PRIORITY VOICE 31.69 1.97 739 HIGH PRIORITY VOICE 0.00 0.00 740 NORMAL PRIORITY DATA 31.22 6.63 741 HIGH PRIORITY DATA 0.00 0.00 742 BEST EFFORT PRIORITY DATA 8.76 11.92 744 Table 4 illustrates the performance of the MAR, MAM, and No-DSTE 745 Bandwidth Constraints Models for a high-day network load pattern with a 746 50% general overload. The numbers given in the table are the total 747 network percent lost (blocked) or delayed traffic. 749 Table 4 750 Performance Comparison for MAR, MAM, & No-DSTE 751 Bandwidth Constraints (BC) Models 752 50% General Overload (Total Network % Lost/Delayed Traffic) 754 Class Type MAR BC MAM BC No-DSTE BC 755 Model Model Model 756 NORMAL PRIORITY VOICE 0.02 0.13 7.98 757 HIGH PRIORITY VOICE 0.00 0.00 8.94 758 NORMAL PRIORITY DATA 0.00 0.26 6.93 759 HIGH PRIORITY DATA 0.00 0.00 8.94 760 BEST EFFORT PRIORITY DATA 10.41 10.39 8.40 762 Again, we can see the performance is always better when MAR bandwidth 763 allocation and reservation is used. 765 Table 5 illustrates the performance of the MAR, MAM, and No-DSTE 766 Bandwidth Constraints Models for a single link failure scenario (3 767 OC-48). The numbers given in the table are the total network percent 768 lost (blocked) or delayed traffic. 770 Table 5 771 Performance Comparison for MAR, MAM, & No-DSTE 772 Bandwidth Constraints (BC) Models 773 Single Link Failure (2 OC-48) 774 (Total Network % Lost/Delayed Traffic) 776 Class Type MAR BC MAM BC No-DSTE BC 777 Model Model Model 778 NORMAL PRIORITY VOICE 0.00 0.62 0.63 779 HIGH PRIORITY VOICE 0.00 0.31 0.32 780 NORMAL PRIORITY DATA 0.00 0.48 0.50 781 HIGH PRIORITY DATA 0.00 0.31 0.32 782 BEST EFFORT PRIORITY DATA 0.12 0.72 0.63 784 Again, we can see the performance is always better when MAR bandwidth 785 allocation and reservation is used. 787 Table 6 illustrates the performance of the MAR, MAM, and No-DSTE 788 Bandwidth Constraints Models for a multiple link failure scenario (3 789 links with 3 OC-48, 3 OC-3, 4 OC-3 capacity, respectively). The numbers 790 Internet Draft MAR Bandwidth Constraints Model for DS-TE Jan 04 792 given in the table are the total network percent lost (blocked) or 793 delayed traffic. 795 Table 6 796 Performance Comparison for MAR, MAM, & No-DSTE 797 Bandwidth Constraints (BC) Models 798 Multiple Link Failure (3 Links with 2 OC-48, 2 OC-12, 1 OC-12, Respectively) 799 (Total Network % Lost/Delayed Traffic) 801 Class Type MAR BC MAM BC No-DSTE BC 802 Model Model Model 803 NORMAL PRIORITY VOICE 0.00 0.91 0.92 804 HIGH PRIORITY VOICE 0.00 0.44 0.44 805 NORMAL PRIORITY DATA 0.00 0.70 0.72 806 HIGH PRIORITY DATA 0.00 0.44 0.44 807 BEST EFFORT PRIORITY DATA 0.14 1.03 1.04 809 Again, we can see the performance is always better when MAR bandwidth 810 allocation and reservation is used. 812 Lai's results [LAI, DSTE-PERF] show the trade-off between bandwidth sharing 813 and service protection/isolation, using an analytic model of a single 814 link. He shows that RDM has a higher degree of sharing than MAM. 815 Furthermore, for a single link, the overall loss probability is the 816 smallest under full sharing and largest under MAM, with RDM being 817 intermediate. Hence, on a single link, Lai shows that the full sharing 818 model yields the highest link efficiency and MAM the lowest, and that 819 full sharing has the poorest service protection capability. 821 The results of the present study show that when considering a network 822 context, in which there are many links and multiple-link routing paths 823 are used, full sharing does not necessarily lead to maximum network-wide 824 bandwidth efficiency. In fact, the results in Table 4 show that the 825 No-DSTE Model not only degrades total network throughput, but also 826 degrades the performance of every CT that should be protected. Allowing 827 more bandwidth sharing may improve performance up to a point, but can 828 severely degrade performance if care is not taken to protect allocated 829 bandwidth under congestion. 831 Both Lai's study and this study show that increasing the degree of 832 bandwidth sharing among the different CTs leads to a tighter coupling 833 between CTs. Under normal loading conditions, there is adequate capacity 834 for each CT, which minimizes the effect of such coupling. Under overload 835 conditions, when there is a scarcity of capacity, such coupling can 836 cause severe degradation of service, especially for the lower priority 837 CTs. 839 Thus, the objective of maximizing efficient bandwidth usage, as stated 840 in Bandwidth Constraints Model objectives, must be exercised with care. 841 Due consideration needs to be given also to achieving bandwidth 842 isolation under overload, in order to minimize the effect of 843 interactions among the different CTs. The proper tradeoff of bandwidth 844 sharing and bandwidth isolation needs to be achieved in the selection of 845 a Bandwidth Constraints Model. Bandwidth reservation supports greater 846 efficiency in bandwidth sharing while still providing bandwidth 847 isolation and protection against QoS degradation. 849 In summary, the proposed MAR Bandwidth Constraints Model includes the 850 following: a) allocate bandwidth to individual CTs, b) protect allocated 851 Internet Draft MAR Bandwidth Constraints Model for DS-TE Jan 04 853 bandwidth by bandwidth reservation methods, as needed, but otherwise 854 fully share bandwidth, c) differentiate high-priority, normal-priority, 855 and best-effort priority services, and d) provide admission control to 856 reject connection requests when needed to meet performance objectives. 858 In the modeling results, the MAR Bandwidth Constraints Model compares 859 favorably with methods that do not use bandwidth reservation. In 860 particular, some of the conclusions from the modeling are as follows: 862 o MAR bandwidth allocation is effective in improving performance over 863 methods that lack bandwidth reservation and that allow more bandwidth 864 sharing under congestion, 865 o MAR achieves service differentiation for high-priority, 866 normal-priority, and best-effort priority services, 867 o bandwidth reservation supports greater efficiency in bandwidth sharing 868 while still providing bandwidth isolation and protection against QoS 869 degradation, and is critical to stable and efficient network 870 performance. 872 Appendix B. Bandwidth Prediction for Path Computation 874 As discussed in [DSTE-PROTO], there there are potential advantages for a 875 Head-end in trying to predict the impact of an LSP on the unreserved 876 bandwidth when computing the path for the LSP. One example would be to 877 perform better load-distribution of multiple LSPs across multiple 878 paths. Another example would be to avoid CAC rejection when the LSP 879 would no longer fit on a link after establishment. 881 Where such predictions are used on Head-ends, the optional Bandwidth 882 Constraints sub-TLV and the optional Maximum Reservable Bandwidth 883 sub-TLV MAY be advertised in the IGP. This can be used by Head-ends 884 to predict how an LSP affects unreserved bandwidth values. Such 885 predictions can be made with MAR by using the unreserved bandwidth 886 values advertised by the IGP, as discussed in Sections 2 and 4: 888 UNRESERVED_BWck = MAX_RESERVABLE_BWk - UNRESERVED_BWk - 889 delta0/1(CTck) * RBW-THRESk 891 where 893 delta0/1(CTck) = 0 if RESERVED_BWck < BCck 894 delta0/1(CTck) = 1 if RESERVED_BWck >= BCck 896 Furthermore, the following estimate can be made for RBW_THRESk: 898 RBW_THRESk = RBW_% * MAX_RESERVABLE_BWk, 900 where RBW_% is a locally configured variable, which could take on 901 different values for different link speeds. This information 902 could be used in conjunction with the BC sub-TLV, 903 MAX_RESERVABLE_BW sub-TLV, and UNRESERVED_BW sub-TLV to make 904 predictions of available bandwidth on each link for each CT. 905 Since admission control algorithms are left for vendor differentiation, 906 predictions can only be performed effectively when the Head-end LSR 907 predictions are based on the same (or a very close) admission control 908 algorithm as used by other LSRs. 910 There may be occasional rejected LSPs when head-ends are establishing 911 LSPs through a common link. As an example, consider some link L, and 912 Internet Draft MAR Bandwidth Constraints Model for DS-TE Jan 04 914 two head-ends H1 and H2. If only H1 or only H2 is establishing LSPs 915 through L, then the prediction is accurate. But, if both H1 and H2 are 916 establishing LSPs through L at the same time, then the prediction 917 would not work perfectly. That is, the CAC will occasionally run into a 918 rejected LSP on a link with such 'race' conditions. Also, as mentioned 919 in Appendix A, such prediction is optional and outside the scope of the 920 document. 922 Full Copyright Statement 924 Copyright (C) The Internet Society (2004). All Rights Reserved. 926 This document and translations of it may be copied and furnished to 927 others, and derivative works that comment on or otherwise explain it or 928 assist in its implementation may be prepared, copied, published and 929 distributed, in whole or in part, without restriction of any kind, 930 provided that the above copyright notice and this paragraph are included 931 on all such copies and derivative works. 933 However, this document itself may not be modified in any way, such as by 934 removing the copyright notice or references to the Internet Society or 935 other Internet organizations, except as needed for the purpose of 936 developing Internet standards in which case the procedures for 937 copyrights defined in the Internet Standards process must be followed, 938 or as required to translate it into languages other than English. 940 The limited permissions granted above are perpetual and will not be 941 revoked by the Internet Society or its successors or assigns. 943 This document and the information contained herein is provided on an "AS 944 IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK 945 FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT 946 LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT 947 INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR 948 FITNESS FOR A PARTICULAR PURPOSE.