idnits 2.17.00 (12 Aug 2021) /tmp/idnits24979/draft-ietf-bier-te-arch-08.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 date (July 13, 2020) is 677 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) -- Looks like a reference, but probably isn't: '2' on line 1333 -- Looks like a reference, but probably isn't: '1' on line 1347 == Missing Reference: 'SI' is mentioned on line 1387, but not defined == Missing Reference: 'I' is mentioned on line 1394, but not defined == Missing Reference: 'VRF' is mentioned on line 1899, but not defined == Unused Reference: 'I-D.qiang-detnet-large-scale-detnet' is defined on line 2054, but no explicit reference was found in the text == Outdated reference: A later version (-06) exists of draft-ietf-bier-multicast-http-response-03 Summary: 0 errors (**), 0 flaws (~~), 6 warnings (==), 3 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group T. Eckert, Ed. 3 Internet-Draft Futurewei 4 Intended status: Standards Track G. Cauchie 5 Expires: January 14, 2021 Bouygues Telecom 6 M. Menth 7 University of Tuebingen 8 July 13, 2020 10 Tree Engineering for Bit Index Explicit Replication (BIER-TE) 11 draft-ietf-bier-te-arch-08 13 Abstract 15 This memo introduces per-packet stateless strict and loose path 16 steered replication and forwarding for Bit Index Explicit Replication 17 packets (RFC8279). This is called BIER Tree Engineering (BIER-TE). 18 BIER-TE can be used as a path steering mechanism in future Traffic 19 Engineering solutions for BIER (BIER-TE). 21 BIER-TE leverages RFC8279 and extends it with a new semantic for bits 22 in the bitstring. BIER-TE can leverage BIER forwarding engines with 23 little or no changes. 25 In BIER, the BitPositions (BP) of the packets bitstring indicate BIER 26 Forwarding Egress Routers (BFER), and hop-by-hop forwarding uses a 27 Routing Underlay such as an IGP. 29 In BIER-TE, BitPositions indicate adjacencies. The BIFT of each BFR 30 are only populated with BPs that are adjacent to the BFR in the BIER- 31 TE topology. The BIER-TE topology can consist of layer 2 or remote 32 (routed) adjacencies. The BFR then replicates and forwards BIER 33 packets to those adjacencies. This results in the aforementioned 34 strict and loose path steering and replications. 36 BIER-TE can co-exist with BIER forwarding in the same domain, for 37 example by using separate BIER sub-domains. In the absence of routed 38 adjacencies, BIER-TE does not require a BIER routing underlay, and 39 can then be operated without requiring an Interior Gateway Routing 40 protocol (IGP). 42 BIER-TE operates without explicit in-network tree-state and carries 43 the multicast distribution tree in the packet header. It can 44 therefore be a good fit to support multicast path steering in Segment 45 Routing (SR) networks. 47 Name explanation 49 [RFC-editor: This section to be removed before publication.] 51 Explanation for name change from BIER-TE to mean "Traffic 52 Engineering" to BIER-TE "Tree Engineering" in WG last-call (to 53 benefit IETF/IESG reviewers): 55 This document started by calling itself BIER-TE, "Traffic 56 Engineering" as it is a mode of BIER specifically beneficial for 57 Traffic Engineering. It supports per-packet bitstring based policy 58 steering and replication. BIER-TE technology itself does not provide 59 a complete traffic engineering solution for BIER but would require 60 combination with other technologies for a full BIER based TE 61 solution, such as a PCE and queuing mechanisms to provide bandwidth 62 and latency reservations. It is also not the only option to build a 63 traffic engineering solution utilizing BIER, for example BIER trees 64 could be steered through IGP metric engineering, such as through 65 Flex-Topologies. The architecure for Traffic Engineering with either 66 modes of BIER (BIER-TE/BIER) is intended to be defined in a separate 67 document, most likely in TEAs WG. 69 Because the name of such an overall solution is intended to be BIER- 70 TE, the expansion of BIER-TE was therefore changed to name this BIER 71 mode "Tree Engineering", so the overall solution can be distinguished 72 better from its tree building/engineering method without having to 73 change the long time well-established abbreviation BIER-TE. 75 Status of This Memo 77 This Internet-Draft is submitted in full conformance with the 78 provisions of BCP 78 and BCP 79. 80 Internet-Drafts are working documents of the Internet Engineering 81 Task Force (IETF). Note that other groups may also distribute 82 working documents as Internet-Drafts. The list of current Internet- 83 Drafts is at https://datatracker.ietf.org/drafts/current/. 85 Internet-Drafts are draft documents valid for a maximum of six months 86 and may be updated, replaced, or obsoleted by other documents at any 87 time. It is inappropriate to use Internet-Drafts as reference 88 material or to cite them other than as "work in progress." 90 This Internet-Draft will expire on January 14, 2021. 92 Copyright Notice 94 Copyright (c) 2020 IETF Trust and the persons identified as the 95 document authors. All rights reserved. 97 This document is subject to BCP 78 and the IETF Trust's Legal 98 Provisions Relating to IETF Documents 99 (https://trustee.ietf.org/license-info) in effect on the date of 100 publication of this document. Please review these documents 101 carefully, as they describe your rights and restrictions with respect 102 to this document. Code Components extracted from this document must 103 include Simplified BSD License text as described in Section 4.e of 104 the Trust Legal Provisions and are provided without warranty as 105 described in the Simplified BSD License. 107 Table of Contents 109 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 4 110 1.1. Basic Examples . . . . . . . . . . . . . . . . . . . . . 5 111 1.2. BIER-TE Topology and adjacencies . . . . . . . . . . . . 8 112 1.3. Comparison with BIER . . . . . . . . . . . . . . . . . . 9 113 1.4. Requirements Language . . . . . . . . . . . . . . . . . . 9 114 2. Components . . . . . . . . . . . . . . . . . . . . . . . . . 9 115 2.1. The Multicast Flow Overlay . . . . . . . . . . . . . . . 10 116 2.2. The BIER-TE Controller . . . . . . . . . . . . . . . . . 10 117 2.2.1. Assignment of BitPositions to adjacencies of the 118 network topology . . . . . . . . . . . . . . . . . . 11 119 2.2.2. Changes in the network topology . . . . . . . . . . . 11 120 2.2.3. Set up per-multicast flow BIER-TE state . . . . . . . 11 121 2.2.4. Link/Node Failures and Recovery . . . . . . . . . . . 12 122 2.3. The BIER-TE Forwarding Layer . . . . . . . . . . . . . . 12 123 2.4. The Routing Underlay . . . . . . . . . . . . . . . . . . 12 124 2.5. Traffic Engineering Considerations . . . . . . . . . . . 12 125 3. BIER-TE Forwarding . . . . . . . . . . . . . . . . . . . . . 13 126 3.1. The Bit Index Forwarding Table (BIFT) . . . . . . . . . . 13 127 3.2. Adjacency Types . . . . . . . . . . . . . . . . . . . . . 14 128 3.2.1. Forward Connected . . . . . . . . . . . . . . . . . . 15 129 3.2.2. Forward Routed . . . . . . . . . . . . . . . . . . . 15 130 3.2.3. ECMP . . . . . . . . . . . . . . . . . . . . . . . . 15 131 3.2.4. Local Decap . . . . . . . . . . . . . . . . . . . . . 16 132 3.3. Encapsulation considerations . . . . . . . . . . . . . . 16 133 3.4. Basic BIER-TE Forwarding Example . . . . . . . . . . . . 16 134 3.5. Forwarding comparison with BIER . . . . . . . . . . . . . 19 135 3.6. Requirements . . . . . . . . . . . . . . . . . . . . . . 19 136 4. BIER-TE Controller BitPosition Assignments . . . . . . . . . 20 137 4.1. P2P Links . . . . . . . . . . . . . . . . . . . . . . . . 20 138 4.2. BFER . . . . . . . . . . . . . . . . . . . . . . . . . . 20 139 4.3. Leaf BFERs . . . . . . . . . . . . . . . . . . . . . . . 20 140 4.4. LANs . . . . . . . . . . . . . . . . . . . . . . . . . . 21 141 4.5. Hub and Spoke . . . . . . . . . . . . . . . . . . . . . . 22 142 4.6. Rings . . . . . . . . . . . . . . . . . . . . . . . . . . 22 143 4.7. Equal Cost MultiPath (ECMP) . . . . . . . . . . . . . . . 23 144 4.8. Routed adjacencies . . . . . . . . . . . . . . . . . . . 26 145 4.8.1. Reducing BitPositions . . . . . . . . . . . . . . . . 26 146 4.8.2. Supporting nodes without BIER-TE . . . . . . . . . . 26 147 4.9. Reuse of BitPositions (without DNR) . . . . . . . . . . . 26 148 4.10. Summary of BP optimizations . . . . . . . . . . . . . . . 28 149 5. Avoiding loops and duplicates . . . . . . . . . . . . . . . . 29 150 5.1. Loops . . . . . . . . . . . . . . . . . . . . . . . . . . 29 151 5.2. Duplicates . . . . . . . . . . . . . . . . . . . . . . . 29 152 6. BIER-TE Forwarding Pseudocode . . . . . . . . . . . . . . . . 29 153 7. Managing SI, subdomains and BFR-ids . . . . . . . . . . . . . 32 154 7.1. Why SI and sub-domains . . . . . . . . . . . . . . . . . 33 155 7.2. Bit assignment comparison BIER and BIER-TE . . . . . . . 34 156 7.3. Using BFR-id with BIER-TE . . . . . . . . . . . . . . . . 34 157 7.4. Assigning BFR-ids for BIER-TE . . . . . . . . . . . . . . 35 158 7.5. Example bit allocations . . . . . . . . . . . . . . . . . 36 159 7.5.1. With BIER . . . . . . . . . . . . . . . . . . . . . . 36 160 7.5.2. With BIER-TE . . . . . . . . . . . . . . . . . . . . 37 161 7.6. Summary . . . . . . . . . . . . . . . . . . . . . . . . . 38 162 8. BIER-TE and Segment Routing . . . . . . . . . . . . . . . . . 38 163 9. Security Considerations . . . . . . . . . . . . . . . . . . . 39 164 10. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 40 165 11. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 40 166 12. Change log [RFC Editor: Please remove] . . . . . . . . . . . 40 167 13. References . . . . . . . . . . . . . . . . . . . . . . . . . 45 168 13.1. Normative References . . . . . . . . . . . . . . . . . . 45 169 13.2. Informative References . . . . . . . . . . . . . . . . . 45 170 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 46 172 1. Introduction 174 BIER-TE shares architecture, terminology and packet formats with BIER 175 as described in [RFC8279] and [RFC8296]. This document describes 176 BIER-TE in the expectation that the reader is familiar with these two 177 documents. 179 In BIER-TE, BitPositions (BP) indicate adjacencies. The BIFT of each 180 BFR is only populated with BP that are adjacent to the BFR in the 181 BIER-TE Topology. Other BPs are left without adjacency. The BFR 182 replicate and forwards BIER packets to adjacent BPs that are set in 183 the packet. BPs are normally also reset upon forwarding to avoid 184 duplicates and loops. This is detailed further below. 186 Note that related work, [I-D.ietf-roll-ccast] uses bloom filters to 187 represent leaves or edges of the intended delivery tree. Bloom 188 filters in general can support larger trees/topologies with fewer 189 addressing bits than explicit bitstrings, but they introduce the 190 heuristic risk of false positives and cannot reset bits in the 191 bitstring during forwarding to avoid loops. For these reasons, BIER- 192 TE uses explicit bitstrings like BIER. The explicit bitstrings of 193 BIER-TE can also be seen as a special type of bloom filter, and this 194 is how related work [ICC] describes it. 196 1.1. Basic Examples 198 BIER-TE forwarding is best introduced with simple examples. 200 BIER-TE Topology: 202 Diagram: 204 p5 p6 205 --- BFR3 --- 206 p3/ p13 \p7 207 BFR1 ---- BFR2 BFR5 ----- BFR6 208 p1 p2 p4\ p14 /p10 p11 p12 209 --- BFR4 --- 210 p8 p9 212 (simplified) BIER-TE Bit Index Forwarding Tables (BIFT): 214 BFR1: p1 -> local_decap 215 p2 -> forward_connected to BFR2 217 BFR2: p1 -> forward_connected to BFR1 218 p5 -> forward_connected to BFR3 219 p8 -> forward_connected to BFR4 221 BFR3: p3 -> forward_connected to BFR2 222 p7 -> forward_connected to BFR5 223 p13 -> local_decap 225 BFR4: p4 -> forward_connected to BFR2 226 p10 -> forward_connected to BFR5 227 p14 -> local_decap 229 BFR5: p6 -> forward_connected to BFR3 230 p9 -> forward_connected to BFR4 231 p12 -> forward_connected to BFR6 233 BFR6: p11 -> forward_connected to BFR5 234 p12 -> local_decap 236 Figure 1: BIER-TE basic example 238 Consider the simple network in the above BIER-TE overview example 239 picture with 6 BFRs. p1...p14 are the BitPositions (BP) used. All 240 BFRs can act as ingress BFR (BFIR), BFR1, BFR3, BFR4 and BFR6 can 241 also be egress BFR (BFER). Forward_connected is the name for 242 adjacencies that are representing subnet adjacencies of the network. 243 Local_decap is the name of the adjacency to decapsulate BIER-TE 244 packets and pass their payload to higher layer processing. 246 Assume a packet from BFR1 should be sent via BFR4 to BFR6. This 247 requires a bitstring (p2,p8,p10,p12). When this packet is examined 248 by BIER-TE on BFR1, the only BitPosition from the bitstring that is 249 also set in the BIFT is p2. This will cause BFR1 to send the only 250 copy of the packet to BFR2. Similarly, BFR2 will forward to BFR4 251 because of p8, BFR4 to BFR5 because of p10 and BFR5 to BFR6 because 252 of p12. p12 also makes BFR6 receive and decapsulate the packet. 254 To send in addition to BFR6 via BFR4 also a copy to BFR3, the 255 bitstring needs to be (p2,p5,p8,p10,p12,p13). When this packet is 256 examined by BFR2, p5 causes one copy to be sent to BFR3 and p8 one 257 copy to BFR4. When BFR3 receives the packet, p13 will cause it to 258 receive and decapsulate the packet. 260 If instead the bitstring was (p2,p6,p8,p10,p12,p13), the packet would 261 be copied by BFR5 towards BFR3 because p6 instead of BFR2 to BFR5 262 because of p6 in the prior case. This is showing the ability of the 263 shown BIER-TE Topology to make the traffic pass across any possible 264 path and be replicated where desired. 266 BIER-TE has various options to minimize BP assignments, many of which 267 are based on assumptions about the required multicast traffic paths 268 and bandwidth consumption in the network. 270 The following picture shows a modified example, in which Rtr2 and 271 Rtr5 are assumed not to support BIER-TE, so traffic has to be unicast 272 encapsulated across them. Unicast tunneling of BIER-TE packets can 273 leverage any feasible mechanism such as MPLS or IP, these 274 encapsulations are out of scope of this document. To emphasize non- 275 native forwarding of BIER-TE packets, these adjacencies are called 276 "forward_routed", but otherwise there is no difference in their 277 processing over the aforementioned "forward_connected" adjacencies. 279 In addition, bits are saved in the following example by assuming that 280 BFR1 only needs to be BFIR but not BFER or transit BFR. 282 BIER-TE Topology: 284 Diagram: 286 p1 p3 p7 287 ....> BFR3 <.... p5 288 ........ ........> 289 BFR1 (Rtr2) (Rtr5) BFR6 290 ........ ........> 291 ....> BFR4 <.... p6 292 p2 p4 p8 294 (simplified) BIER-TE Bit Index Forwarding Tables (BIFT): 296 BFR1: p1 -> forward_routed to BFR3 297 p2 -> forward_routed to BFR4 299 BFR3: p3 -> local_decap 300 p5 -> forward_routed to BFR6 302 BFR4: p4 -> local_decap 303 p6 -> forward_routed to BFR6 305 BFR6: p5 -> local_decap 306 p6 -> local_decap 307 p7 -> forward_routed to BFR3 308 p8 -> forward_routed to BFR4 310 Figure 2: BIER-TE basic overlay example 312 To send a BIER-TE packet from BFR1 via BFR3 to BFR6, the bitstring is 313 (p1,p5). From BFR1 via BFR4 to BFR6 it is (p2,p6). A packet from 314 BFR1 to BFR3,BFR4 and BFR6 can use (p1,p2,p3,p4,p5) or 315 (p1,p2,p3,p4,p6), or via BFR6 (p2,p3,p4,p6,p7) or (p1.p3,p4,p5,p8). 317 1.2. BIER-TE Topology and adjacencies 319 The key new component in BIER-TE to control where replication can or 320 should happens and how to minimize the required BP for segments is - 321 as shown in these two examples - the BIER-TE topology. 323 The BIER-TE Topology consists of the BIFT of all the BFR and can also 324 be expressed in a diagram as a graph where the edges are the 325 adjacencies between the BFR. Adjacencies are naturally 326 unidirectional. BP can be reused across multiple adjacencies as long 327 as this does not lead to undesired duplicates or loops as explained 328 further down in the text. 330 If the BIER-TE topology represents the underlying (layer 2) topology 331 of the network, this is called "native" BIER-TE as shown in the first 332 example. This can be freely mixed with "overlay" BIER-TE, in 333 "forward_routed" adjacencies are used. 335 1.3. Comparison with BIER 337 The key differences over BIER are: 339 o BIER-TE replaces in-network autonomous path calculation by 340 explicit paths calculated off-path by the BIER-TE Controller. 342 o In BIER-TE every BitPosition of the BitString of a BIER-TE packet 343 indicates one or more adjacencies - instead of a BFER as in BIER. 345 o BIER-TE in each BFR has no routing table but only a BIER-TE 346 Forwarding Table (BIFT) indexed by SI:BitPosition and populated 347 with only those adjacencies to which the BFR should replicate 348 packets to. 350 BIER-TE headers use the same format as BIER headers. 352 BIER-TE forwarding does not require/use the BFIR-ID. The BFIR-ID can 353 still be useful though for coordinated BFIR/BFER functions, such as 354 the context for upstream assigned labels for MPLS payloads in MVPN 355 over BIER-TE. 357 If the BIER-TE domain is also running BIER, then the BFIR-ID in BIER- 358 TE packets can be set to the same BFIR-ID as used with BIER packets. 360 If the BIER-TE domain is not running full BIER or does not want to 361 reduce the need to allocate bits in BIER bitstrings for BFIR-ID 362 values, then the allocation of BFIR-ID values in BIER-TE packets can 363 be done through other mechanisms outside the scope of this document, 364 as long as this is appropriately agreed upon between all BFIR/BFER. 366 1.4. Requirements Language 368 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 369 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 370 document are to be interpreted as described in RFC 2119 [RFC2119]. 372 2. Components 374 End to end BIER-TE operations consists of four mayor components: The 375 "Multicast Flow Overlay", the "BIER-TE control plane" consisting of 376 the "BIER-TE Controller" and its signaling channels to the BFR, the 377 "Routing Underlay" and the "BIER-TE forwarding layer". The Bier-TE 378 Controller is the new architectural component in BIER-TE compared to 379 BIER. 381 Picture 2: Components of BIER-TE 383 <------BGP/PIM-----> 384 |<-IGMP/PIM-> multicast flow <-PIM/IGMP->| 385 overlay 387 [BIER-TE Controller] <=> [BIER-TE Topology] 388 BIER-TE control plane 389 ^ ^ ^ 390 / | \ BIER-TE control protocol 391 | | | e.g. Netconf/Restconf/Yang 392 v v v 393 Src -> Rtr1 -> BFIR-----BFR-----BFER -> Rtr2 -> Rcvr 395 |<----------------->| 396 BIER-TE forwarding layer 398 |<- BIER-TE domain->| 400 |<--------------------->| 401 Routing underlay 403 Figure 3: BIER-TE architecture 405 2.1. The Multicast Flow Overlay 407 The Multicast Flow Overlay operates as in BIER. See [RFC8279]. 408 Instead of interacting with the BIER forwarding layer (as in BIER), 409 it interacts with the BIER-TE Controller. 411 2.2. The BIER-TE Controller 413 The BIER-TE Controller is representing the control plane of BIER-TE. 414 It communicates two sets of information with BFRs: 416 During initial provisioning or modifications of the network topology, 417 the BIER-TE Controller discovers the network topology and creates the 418 BIER-TE topology from it: determine which adjacencies are required/ 419 desired and assign BitPositions to them. Then it signals the 420 resulting of BitPositions and their adjacencies to each BFR to set up 421 their BIER-TE BIFTs. 423 During day-to-day operations of the network, the BIER-TE Controller 424 signals to BFIRs what multicast flows are mapped to what BitStrings. 426 Communications between the BIER-TE Controller and BFRs is ideally via 427 standardized protocols and data-models such as Netconf/Restconf/Yang. 428 This is currently outside the scope of this document. Vendor- 429 specific CLI on the BFRs is also a possible stopgap option (as in 430 many other SDN solutions lacking definition of standardized data 431 model). 433 For simplicity, the procedures of the BIER-TE Controller are 434 described in this document as if it is a single, centralized 435 automated entity, such as an SDN controller. It could equally be an 436 operator setting up CLI on the BFRs. Distribution of the functions 437 of the BIER-TE Controller is currently outside the scope of this 438 document. 440 2.2.1. Assignment of BitPositions to adjacencies of the network 441 topology 443 The BIER-TE Controller tracks the BFR topology of the BIER-TE domain. 444 It determines what adjacencies require BitPositions so that BIER-TE 445 explicit paths can be built through them as desired by operator 446 policy. 448 The BIER-TE Controller then pushes the BitPositions/adjacencies to 449 the BIFT of the BFRs, populating only those SI:BitPositions to the 450 BIFT of each BFR to which that BFR should be able to send packets to 451 - adjacencies connecting to this BFR. 453 2.2.2. Changes in the network topology 455 If the network topology changes (not failure based) so that 456 adjacencies that are assigned to BitPositions are no longer needed, 457 the BIER-TE Controller can re-use those BitPositions for new 458 adjacencies. First, these BitPositions need to be removed from any 459 BFIR flow state and BFR BIFT state, then they can be repopulated, 460 first into BIFT and then into the BFIR. 462 2.2.3. Set up per-multicast flow BIER-TE state 464 The BIER-TE Controller interacts with the multicast flow overlay to 465 determine what multicast flow needs to be sent by a BFIR to which set 466 of BFER. It calculates the desired distribution tree across the 467 BIER-TE domain based on algorithms outside the scope of this document 468 (e.g. CSFP, Steiner Tree, ...). It then pushes the calculated 469 BitString into the BFIR. 471 See [I-D.ietf-bier-multicast-http-response] for a solution describing 472 this interaction. 474 2.2.4. Link/Node Failures and Recovery 476 When link or nodes fail or recover in the topology, BIER-TE can 477 quickly respond with the optional FRR procedures described in [I- 478 D.eckert-bier-te-frr]. It can also more slowly react by 479 recalculating the BitStrings of affected multicast flows. This 480 reaction is slower than the FRR procedure because the BIER-TE 481 Controller needs to receive link/node up/down indications, 482 recalculate the desired BitStrings and push them down into the BFIRs. 483 With FRR, this is all performed locally on a BFR receiving the 484 adjacency up/down notification. 486 2.3. The BIER-TE Forwarding Layer 488 When the BIER-TE Forwarding Layer receives a packet, it simply looks 489 up the BitPositions that are set in the BitString of the packet in 490 the Bit Index Forwarding Table (BIFT) that was populated by the BIER- 491 TE Controller. For every BP that is set in the BitString, and that 492 has one or more adjacencies in the BIFT, a copy is made according to 493 the type of adjacencies for that BP in the BIFT. Before sending any 494 copy, the BFR resets all BP in the BitString of the packet for which 495 the BFR has one or more adjacencies in the BIFT, except when the 496 adjacency indicates "DoNotReset" (DNR, see Section 3.2.1). This is 497 done to inhibit that packets can loop. 499 2.4. The Routing Underlay 501 BIER-TE is sending BIER packets to directly connected BIER-TE 502 neighbors as L2 (unicasted) BIER packets without requiring a routing 503 underlay. BIER-TE forwarding uses the Routing underlay for 504 forward_routed adjacencies which copy BIER-TE packets to not- 505 directly-connected BFRs (see below for adjacency definitions). 507 If the BFR intends to support FRR for BIER-TE, then the BIER-TE 508 forwarding plane needs to receive fast adjacency up/down 509 notifications: Link up/down or neighbor up/down, e.g. from BFD. 510 Providing these notifications is considered to be part of the routing 511 underlay in this document. 513 2.5. Traffic Engineering Considerations 515 Traffic Engineering ([I-D.dt-teas-rfc3272bis]) provides performance 516 optimization of operational IP networks while utilizing network 517 resources economically and reliably. The key elements needed to 518 effect TE are policy, path steering and resource management. These 519 elements require support at the control/controller level and within 520 the forwarding plane. 522 Policy decisions are made within the BIER-TE control plane, i.e., 523 within BIER-TE Controllers. Controllers use policy when composing 524 BitStrings (BFR flow state) and BFR BIFT state. The mapping of user/ 525 IP traffic to specific BitStrings/BIER-TE flows is made based on 526 policy. The specifics details of BIER-TE policies and how a 527 controller uses such are out of scope of this document. 529 Path steering is supported via the definition of a BitString. 530 BitStrings used in BIER-TE are composed based on policy and resource 531 management considerations. When composing BIER-TE BitStrings, a 532 Controller MUST take into account the resources available at each BFR 533 and for each BP when it is providing congestion loss free services 534 such as for Rate Controlled Service Disciplines (RCSD). Resource 535 availability could be provided for example via routing protocol 536 information, but may also be obtained via a BIER-TE control protocol. 537 The resource usage of the BIER-TE traffic admitted by the BIER-TE 538 controller can be solely tracked on the BIER-TE Controller based on 539 local accounting as long as no forward_routed adjacencies are used 540 (see below for definition of forward_routed adjacencies). When 541 forward_routed adjacencies are used, the paths selected by the 542 underlying routing protocol need to be tracked as well. 544 Resource management has implications on the forwarding plane beyond 545 the BIER-TE defined steering of packets. This includes allocation of 546 buffers to guarantee the worst case requirements of admitted RCSD 547 trafic and potential policing and/or rate-shaping mechanisms, 548 typically done via various forms of queuing. This level of resource 549 control, while optional, is important in networks that wish to 550 support congestion management policies to control or regulate the 551 offered traffic to deliver different levels of service and alleviate 552 congestion problems, or those networks that wish to control latencies 553 experienced by specific traffic flows. 555 3. BIER-TE Forwarding 557 3.1. The Bit Index Forwarding Table (BIFT) 559 The Bit Index Forwarding Table (BIFT) exists in every BFR. For every 560 subdomain in use, it is a table indexed by SI:BitPosition and is 561 populated by the BIER-TE control plane. Each index can be empty or 562 contain a list of one or more adjacencies. 564 BIER-TE can support multiple subdomains like BIER. Each one with a 565 separate BIFT 567 In the BIER architecture, indices into the BIFT are explained to be 568 both BFR-id and SI:BitString (BitPosition). This is because there is 569 a 1:1 relationship between BFR-id and SI:BitString - every bit in 570 every SI is/can be assigned to a BFIR/BFER. In BIER-TE there are 571 more bits used in each BitString than there are BFIR/BFER assigned to 572 the bitstring. This is because of the bits required to express the 573 engineered path through the topology. The BIER-TE forwarding 574 definitions do therefore not use the term BFR-id at all. Instead, 575 BFR-ids are only used as required by routing underlay, flow overlay 576 of BIER headers. Please refer to Section 7 for explanations how to 577 deal with SI, subdomains and BFR-id in BIER-TE. 579 ------------------------------------------------------------------ 580 | Index: | Adjacencies: | 581 | SI:BitPosition | or one or more per entry | 582 ================================================================== 583 | 0:1 | forward_connected(interface,neighbor{,DNR}) | 584 ------------------------------------------------------------------ 585 | 0:2 | forward_connected(interface,neighbor{,DNR}) | 586 | | forward_connected(interface,neighbor{,DNR}) | 587 ------------------------------------------------------------------ 588 | 0:3 | local_decap({VRF}) | 589 ------------------------------------------------------------------ 590 | 0:4 | forward_routed({VRF,}l3-neighbor) | 591 ------------------------------------------------------------------ 592 | 0:5 | | 593 ------------------------------------------------------------------ 594 | 0:6 | ECMP({adjacency1,...adjacencyN}, seed) | 595 ------------------------------------------------------------------ 596 ... 597 | BitStringLength | ... | 598 ------------------------------------------------------------------ 599 Bit Index Forwarding Table 601 Figure 4: BIFT adjacencies 603 The BIFT is programmed into the data plane of BFRs by the BIER-TE 604 Controller and used to forward packets, according to the rules 605 specified in the BIER-TE Forwarding Procedures. 607 Adjacencies for the same BP when populated in more than one BFR by 608 the BIER-TE Controller does not have to have the same adjacencies. 609 This is up to the BIER-TE Controller. BPs for p2p links are one case 610 (see below). 612 3.2. Adjacency Types 613 3.2.1. Forward Connected 615 A "forward_connected" adjacency is towards a directly connected BFR 616 neighbor using an interface address of that BFR on the connecting 617 interface. A forward_connected adjacency does not route packets but 618 only L2 forwards them to the neighbor. 620 Packets sent to an adjacency with "DoNotReset" (DNR) set in the BIFT 621 will not have the BitPosition for that adjacency reset when the BFR 622 creates a copy for it. The BitPosition will still be reset for 623 copies of the packet made towards other adjacencies. This can be 624 used for example in ring topologies as explained below. 626 3.2.2. Forward Routed 628 A "forward_routed" adjacency is an adjacency towards a BFR that is 629 not a forward_connected adjacency: towards a loopback address of a 630 BFR or towards an interface address that is non-directly connected. 631 Forward_routed packets are forwarded via the Routing Underlay. 633 If the Routing Underlay has multiple paths for a forward_routed 634 adjacency, it will perform ECMP independent of BIER-TE for packets 635 forwarded across a forward_routed adjacency. This is independent of 636 BIER-TE ECMP described in Section 3.2.3. 638 If the Routing Underlay has FRR, it will perform FRR independent of 639 BIER-TE for packets forwarded across a forward_routed adjacency. 641 3.2.3. ECMP 643 The ECMP mechanisms in BIER are tied to the BIER BIFT and are 644 therefore not directly useable with BIER-TE. The following 645 procedures describe ECMP for BIER-TE that we consider to be 646 lightweight but also well manageable. It leverages the existing 647 entropy parameter in the BIER header to keep packets of the flows on 648 the same path and it introduces a "seed" parameter to allow for 649 traffic flows to be polarized or randomized across multiple hops. 651 An "Equal Cost Multipath" (ECMP) adjacency has a list of two or more 652 adjacencies included in it. It copies the BIER-TE to one of those 653 adjacencies based on the ECMP hash calculation. The BIER-TE ECMP 654 hash algorithm must select the same adjacency from that list for all 655 packets with the same "entropy" value in the BIER-TE header if the 656 same number of adjacencies and same seed are given as parameters. 657 Further use of the seed parameter is explained below. 659 3.2.4. Local Decap 661 A "local_decap" adjacency passes a copy of the payload of the BIER-TE 662 packet to the packets NextProto within the BFR (IPv4/IPv6, 663 Ethernet,...). A local_decap adjacency turns the BFR into a BFER for 664 matching packets. Local_decap adjacencies require the BFER to 665 support routing or switching for NextProto to determine how to 666 further process the packet. 668 3.3. Encapsulation considerations 670 Specifications for BIER-TE encapsulation are outside the scope of 671 this document. This section gives explanations and guidelines. 673 Because a BFR needs to interpret the BitString of a BIER-TE packet 674 differently from a BIER packet, it is necessary to distinguish BIER 675 from BIER-TE packets. This is subject to definitions in BIER 676 encapsulation specifications. 678 MPLS encapsulation [RFC8296] for example assigns one label by which 679 BFRs recognizes BIER packets for every (SI,subdomain) combination. 680 If it is desirable that every subdomain can forward only BIER or 681 BIER-TE packets, then the label allocation could stay the same, and 682 only the forwarding model (BIER/BIER-TE) would have to be defined per 683 subdomain. If it is desirable to support both BIER and BIER-TE 684 forwarding in the same subdomain, then additional labels would need 685 to be assigned for BIER-TE forwarding. 687 "forward_routed" requires an encapsulation permitting to unicast 688 BIER-TE packets to a specific interface address on a target BFR. 689 With MPLS encapsulation, this can simply be done via a label stack 690 with that addresses label as the top label - followed by the label 691 assigned to (SI,subdomain) - and if necessary (see above) BIER-TE. 692 With non-MPLS encapsulation, some form of IP tunneling (IP in IP, 693 LISP, GRE) would be required. 695 The encapsulation used for "forward_routed" adjacencies can equally 696 support existing advanced adjacency information such as "loose source 697 routes" via e.g. MPLS label stacks or appropriate header extensions 698 (e.g. for IPv6). 700 3.4. Basic BIER-TE Forwarding Example 702 [RFC Editor: remove this section.] 704 THIS SECTION TO BE REMOVED IN RFC BECAUSE IT WAS SUPERCEEDED BY 705 SECTION 1.1 EXAMPLE - UNLESS REVIEWERS CHIME IN AND EXPRESS DESIRE TO 706 KEEP THIS ADDITIONAL EXAMPLE SECTION. 708 Step by step example of basic BIER-TE forwarding. This does not use 709 ECMP or forward_routed adjacencies nor does it try to minimize the 710 number of required BitPositions for the topology. 712 [BIER-TE Controller] 713 / | \ 714 v v v 716 | p13 p1 | 717 +- BFIR2 --+ | 718 | | p2 p6 | LAN2 719 | +-- BFR3 --+ | 720 | | | p7 p11 | 721 Src -+ +-- BFER1 --+ 722 | | p3 p8 | | 723 | +-- BFR4 --+ +-- Rcv1 724 | | | | 725 | | 726 | p14 p4 | 727 +- BFIR1 --+ | 728 | +-- BFR5 --+ p10 p12 | 729 LAN1 | p5 p9 +-- BFER2 --+ 730 | +-- Rcv2 731 | 732 LAN3 734 IP |..... BIER-TE network......| IP 736 Figure 5: BIER-TE Forwarding Example 738 pXX indicate the BitPositions number assigned by the BIER-TE 739 Controller to adjacencies in the BIER-TE topology. For example, p9 740 is the adjacency towards BFR5 on the LAN connecting to BFER2. 742 BIFT BFIR2: 743 p13: local_decap() 744 p2: forward_connected(BFR3) 746 BIFT BFR3: 747 p1: forward_connected(BFIR2) 748 p7: forward_connected(BFER1) 749 p8: forward_connected(BFR4) 751 BIFT BFER1: 752 p11: local_decap() 753 p6: forward_connected(BFR3) 754 p8: forward_connected(BFR4) 756 Figure 6: BIER-TE Forwarding Example Adjacencies 758 ...and so on. 760 For example, we assume that some multicast traffic seen on LAN1 needs 761 to be sent via BIER-TE by BFIR2 towards Rcv1 and Rcv2. The BIER-TE 762 Controller determines it wants it to pass this traffic across the 763 following paths: 765 -> BFER1 ---------------> Rcv1 766 BFIR2 -> BFR3 767 -> BFR4 -> BFR5 -> BFER2 -> Rcv2 769 Figure 7: BIER-TE Forwarding Example Paths 771 These paths equal to the following BitString: p2, p5, p7, p8, p10, 772 p11, p12. 774 This BitString is assigned by BFIR2 to the example multicast traffic 775 received from LAN1. 777 Then BFIR2 forwards this multicast traffic with BIER-TE based on that 778 BitString. The BIFT of BFIR2 has only p2 and p13 populated. Only p2 779 is in the BitString and this is an adjacency towards BFR3. BFIR2 780 therefore resets p2 in the BitString and sends a copy towards BFR2. 782 BFR3 sees a BitString of p5,p7,p8,p10,p11,p12. It is only interested 783 in p1,p7,p8. It creates a copy of the packet to BFER1 (due to p7) 784 and one to BFR4 (due to p8). It resets p7, p8 before sending. 786 BFER1 sees a BitString of p5,p10,p11,p12. It is only interested in 787 p6,p7,p8,p11 and therefore considers only p11. p11 is a "local_decap" 788 adjacency installed by the BIER-TE Controller because BFER1 should 789 pass packets to IP multicast. The local_decap adjacency instructs 790 BFER1 to create a copy, decapsulate it from the BIER header and pass 791 it on to the NextProtocol, in this example IP multicast. IP 792 multicast will then forward the packet out to LAN2 because it did 793 receive PIM or IGMP joins on LAN2 for the traffic. 795 Further processing of the packet in BFR4, BFR5 and BFER2 accordingly. 797 3.5. Forwarding comparison with BIER 799 Forwarding of BIER-TE is designed to allow common forwarding hardware 800 with BIER. In fact, one of the main goals of this document is to 801 encourage the building of forwarding hardware that can not only 802 support BIER, but also BIER-TE - to allow experimentation with BIER- 803 TE and support building of BIER-TE control plane code. 805 The pseudocode in Section 6 shows how existing BIER/BIFT forwarding 806 can be amended to support basic BIER-TE forwarding, by using BIER 807 BIFT's F-BM. Only the masking of bits due to avoid duplicates must 808 be skipped when forwarding is for BIER-TE. 810 Whether to use BIER or BIER-TE forwarding can simply be a configured 811 choice per subdomain and accordingly be set up by a BIER-TE 812 Controller. The BIER packet encapsulation [RFC8296] too can be 813 reused without changes except that the currently defined BIER-TE ECMP 814 adjacency does not leverage the entropy field so that field would be 815 unused when BIER-TE forwarding is used. 817 3.6. Requirements 819 Basic BIER-TE forwarding MUST support to configure Subdomains to use 820 basic BIER-TE forwarding rules (instead of BIER). With basic BIER-TE 821 forwarding, every bit MUST support to have zero or one adjacency. It 822 MUST support the adjacency types forward_connected without DNR flag, 823 forward_routed and local_decap. All other BIER-TE forwarding 824 features are optional. These basic BIER-TE requirements make BIER-TE 825 forwarding exactly the same as BIER forwarding with the exception of 826 skipping the aforementioned F-BM masking on egress. 828 BIER-TE forwarding SHOULD support the DNR flag, as this is highly 829 useful to save bits in rings (see Section 4.6). 831 BIER-TE forwarding MAY support more than one adjacency on a bit and 832 ECMP adjacencies. The importance of ECMP adjacencies is unclear when 833 traffic steering is used because it may be more desirable to 834 explicitly steer traffic across non-ECMP paths to make per-path 835 traffic calculation easier for BIER-TE Controllers. Having more than 836 one adjacency for a bit allows further savings of bits in hub&spoke 837 scenarios, but unlike rings it is less "natural" to flood traffic 838 across multiple links unconditional. Both ECMP and multiple 839 adjacencies are forwarding plane features that should be possible to 840 support later when needed as they do not impact the basic BIER-TE 841 replication loop. This is true because there is no inter-copy 842 dependency through resetting of F-BM as in BIER. 844 4. BIER-TE Controller BitPosition Assignments 846 This section describes how the BIER-TE Controller can use the 847 different BIER-TE adjacency types to define the BitPositions of a 848 BIER-TE domain. 850 Because the size of the BitString is limiting the size of the BIER-TE 851 domain, many of the options described exist to support larger 852 topologies with fewer BitPositions (4.1, 4.3, 4.4, 4.5, 4.6, 4.7, 853 4.8). 855 4.1. P2P Links 857 Each P2p link in the BIER-TE domain is assigned one unique 858 BitPosition with a forward_connected adjacency pointing to the 859 neighbor on the p2p link. 861 4.2. BFER 863 Every non-Leaf BFER is given a unique BitPosition with a local_decap 864 adjacency. 866 4.3. Leaf BFERs 868 BFR1(P) BFR2(P) BFR1(P) BFR2(P) 869 | \ / | | | 870 | X | | | 871 | / \ | | | 872 BFER1(PE) BFER2(PE) BFER1(PE)----BFER2(PE) 874 Leaf BFER / Non-Leaf BFER / 875 PE-router PE-router 877 Figure 8: Leaf vs. non-Leaf BFER Example 879 Leaf BFERs are BFERs where incoming BIER-TE packets never need to be 880 forwarded to another BFR but are only sent to the BFER to exit the 881 BIER-TE domain. For example, in networks where PEs are spokes 882 connected to P routers, those PEs are Leaf BFERs unless there is a 883 U-turn between two PEs. Consider how redundant disjoint traffic can 884 reach BFER1/BFER2 in above picture: When BFER1/BFER2 are Non-Leaf 885 BFER as shown on the right hand side, one traffic copy would be 886 forwarded to BFER1 from BFR1, but the other one could only reach 887 BFER1 via BFER2, which makes BFER2 a non-Leaf BFER. Likewise BFER1 888 is a non-Leaf BFER when forwarding traffic to BFER2. 890 Note that the BFERs in the left hand picture are only guaranteed to 891 be leaf-BFER by fitting routing configuration that prohibits transit 892 traffic to pass through the BFERs, which is commonly applied in these 893 topologies. 895 All leaf-BFER in a BIER-TE domain can share a single BitPosition. 896 This is possible because the BitPosition for the adjacency to reach 897 the BFER can be used to distinguish whether or not packets should 898 reach the BFER. 900 This optimization will not work if an upstream interface of the BFER 901 is using a BitPosition optimized as described in the following two 902 sections (LAN, Hub and Spoke). 904 4.4. LANs 906 In a LAN, the adjacency to each neighboring BFR on the LAN is given a 907 unique BitPosition. The adjacency of this BitPosition is a 908 forward_connected adjacency towards the BFR and this BitPosition is 909 populated into the BIFT of all the other BFRs on that LAN. 911 BFR1 912 |p1 913 LAN1-+-+---+-----+ 914 p3| p4| p2| 915 BFR3 BFR4 BFR7 917 Figure 9: LAN Example 919 If Bandwidth on the LAN is not an issue and most BIER-TE traffic 920 should be copied to all neighbors on a LAN, then BitPositions can be 921 saved by assigning just a single BitPosition to the LAN and 922 populating the BitPosition of the BIFTs of each BFRs on the LAN with 923 a list of forward_connected adjacencies to all other neighbors on the 924 LAN. 926 This optimization does not work in the case of BFRs redundantly 927 connected to more than one LANs with this optimization because these 928 BFRs would receive duplicates and forward those duplicates into the 929 opposite LANs. Adjacencies of such BFRs into their LANs still need a 930 separate BitPosition. 932 4.5. Hub and Spoke 934 In a setup with a hub and multiple spokes connected via separate p2p 935 links to the hub, all p2p links can share the same BitPosition. The 936 BitPosition on the hub's BIFT is set up with a list of 937 forward_connected adjacencies, one for each Spoke. 939 This option is similar to the BitPosition optimization in LANs: 940 Redundantly connected spokes need their own BitPositions. 942 This type of optimized BP could be used for example when all traffic 943 is "broadcast" traffic (very dense receiver set) such as live-TV or 944 situation-awareness (SA). This BP optimization can then be used to 945 explicitly steer different traffic flows across different ECMP paths 946 in Data-Center or broadband-aggregation networks with minimal use of 947 BPs. 949 4.6. Rings 951 In L3 rings, instead of assigning a single BitPosition for every p2p 952 link in the ring, it is possible to save BitPositions by setting the 953 "Do Not Reset" (DNR) flag on forward_connected adjacencies. 955 For the rings shown in the following picture, a single BitPosition 956 will suffice to forward traffic entering the ring at BFRa or BFRb all 957 the way up to BFR1: 959 On BFRa, BFRb, BFR30,... BFR3, the BitPosition is populated with a 960 forward_connected adjacency pointing to the clockwise neighbor on the 961 ring and with DNR set. On BFR2, the adjacency also points to the 962 clockwise neighbor BFR1, but without DNR set. 964 Handling DNR this way ensures that copies forwarded from any BFR in 965 the ring to a BFR outside the ring will not have the ring BitPosition 966 set, therefore minimizing the chance to create loops. 968 v v 969 | | 970 L1 | L2 | L3 971 /-------- BFRa ---- BFRb --------------------\ 972 | | 973 \- BFR1 - BFR2 - BFR3 - ... - BFR29 - BFR30 -/ 974 | | L4 | | 975 p33| p15| 976 BFRd BFRc 978 Figure 10: Ring Example 980 Note that this example only permits for packets to enter the ring at 981 BFRa and BFRb, and that packets will always travel clockwise. If 982 packets should be allowed to enter the ring at any ring BFR, then one 983 would have to use two ring BitPositions. One for clockwise, one for 984 counterclockwise. 986 Both would be set up to stop rotating on the same link, e.g. L1. 987 When the ingress ring BFR creates the clockwise copy, it will reset 988 the counterclockwise BitPosition because the DNR bit only applies to 989 the bit for which the replication is done. Likewise for the 990 clockwise BitPosition for the counterclockwise copy. In result, the 991 ring ingress BFR will send a copy in both directions, serving BFRs on 992 either side of the ring up to L1. 994 4.7. Equal Cost MultiPath (ECMP) 996 The ECMP adjacency allows to use just one BP per link bundle between 997 two BFRs instead of one BP for each p2p member link of that link 998 bundle. In the following picture, one BP is used across L1,L2,L3. 1000 --L1----- 1001 BFR1 --L2----- BFR2 1002 --L3----- 1004 BIFT entry in BFR1: 1005 ------------------------------------------------------------------ 1006 | Index | Adjacencies | 1007 ================================================================== 1008 | 0:6 | ECMP({forward_connected(L1, BFR2), | 1009 | | forward_connected(L2, BFR2), | 1010 | | forward_connected(L3, BFR2)}, seed) | 1011 ------------------------------------------------------------------ 1013 BIFT entry in BFR2: 1014 ------------------------------------------------------------------ 1015 | Index | Adjacencies | 1016 ================================================================== 1017 | 0:6 | ECMP({forward_connected(L1, BFR1), | 1018 | | forward_connected(L2, BFR1), | 1019 | | forward_connected(L3, BFR1)}, seed) | 1020 ------------------------------------------------------------------ 1022 Figure 11: ECMP Example 1024 This document does not standardize any ECMP algorithm because it is 1025 sufficient for implementations to document their freely chosen ECMP 1026 algorithm. This allows the BIER-TE Controller to calculate ECMP 1027 paths and seeds. The following picture shows an example ECMP 1028 algorithm: 1030 forward(packet, ECMP(adj(0), adj(1),... adj(N-1), seed)): 1031 i = (packet(bier-header-entropy) XOR seed) % N 1032 forward packet to adj(i) 1034 Figure 12: ECMP algorithm Example 1036 In the following example, all traffic from BFR1 towards BFR10 is 1037 intended to be ECMP load split equally across the topology. This 1038 example is not meant as a likely setup, but to illustrate that ECMP 1039 can be used to share BPs not only across link bundles, and it 1040 explains the use of the seed parameter. 1042 BFR1 (BFIR) 1043 /L11 \L12 1044 / \ 1045 BFR2 BFR3 1046 /L21 \L22 /L31 \L32 1047 / \ / \ 1048 BFR4 BFR5 BFR6 BFR7 1049 \ / \ / 1050 \ / \ / 1051 BFR8 BFR9 1052 \ / 1053 \ / 1054 BFR10 (BFER) 1056 BIFT entry in BFR1: 1057 ------------------------------------------------------------------ 1058 | 0:6 | ECMP({forward_connected(L11, BFR2), | 1059 | | forward_connected(L12, BFR3)}, seed1) | 1060 ------------------------------------------------------------------ 1062 BIFT entry in BFR2: 1063 ------------------------------------------------------------------ 1064 | 0:7 | ECMP({forward_connected(L21, BFR4), | 1065 | | forward_connected(L22, BFR5)}, seed1) | 1066 ------------------------------------------------------------------ 1068 BIFT entry in BFR3: 1069 ------------------------------------------------------------------ 1070 | 0:7 | ECMP({forward_connected(L31, BFR6), | 1071 | | forward_connected(L32, BFR7)}, seed1) | 1072 ------------------------------------------------------------------ 1073 BIFT entry in BFR4, BFR5: 1074 ------------------------------------------------------------------ 1075 | 0:8 | forward_connected(Lxx, BFR8) |xx differs on BFR4/BFR5| 1076 ------------------------------------------------------------------ 1078 BIFT entry in BFR6, BFR7: 1079 ------------------------------------------------------------------ 1080 | 0:8 | forward_connected(Lxx, BFR9) |xx differs on BFR6/BFR7| 1081 ------------------------------------------------------------------ 1083 BIFT entry in BFR8, BFR9: 1084 ------------------------------------------------------------------ 1085 | 0:9 | forward_connected(Lxx, BFR10) |xx differs on BFR8/BFR9| 1086 ------------------------------------------------------------------ 1088 Figure 13: Polarization Example 1090 Note that for the following discussion of ECMP, only the BIFT ECMP 1091 adjacencies on BFR1, BFR2, BFR3 are relevant. The re-use of BP 1092 across BFR in this example is further explained in Section 4.9 below. 1094 With the setup of ECMP in above topology, traffic would not be 1095 equally load-split. Instead, links L22 and L31 would see no traffic 1096 at all: BFR2 will only see traffic from BFR1 for which the ECMP hash 1097 in BFR1 selected the first adjacency in the list of 2 adjacencies 1098 given as parameters to the ECMP. It is link L11-to-BFR2. BFR2 1099 performs again ECMP with two adjacencies on that subset of traffic 1100 using the same seed1, and will therefore again select the first of 1101 its two adjacencies: L21-to-BFR4. And therefore L22 and BFR5 sees no 1102 traffic. Likewise for L31 and BFR6. 1104 This issue in BFR2/BFR3 is called polarization. It results from the 1105 re-use of the same hash function across multiple consecutive hops in 1106 topologies like these. To resolve this issue, the ECMP adjacency on 1107 BFR1 can be set up with a different seed2 than the ECMP adjacencies 1108 on BFR2/BFR3. BFR2/BFR3 can use the same hash because packets will 1109 not sequentially pass across both of them. Therefore, they can also 1110 use the same BP 0:7. 1112 Note that ECMP solutions outside of BIER often hide the seed by auto- 1113 selecting it from local entropy such as unique local or next-hop 1114 identifiers. The solutions choosen for BIER-TE to allow the BIER-TE 1115 Controller to explicitly set the seed maximizes the ability of the 1116 BIER-TE Controller to choose the seed, independent of such seed 1117 source that the BIER-TE Controller may not be able to control well, 1118 and even calculate optimized seeds for multi-hop cases. 1120 4.8. Routed adjacencies 1122 4.8.1. Reducing BitPositions 1124 Routed adjacencies can reduce the number of BitPositions required 1125 when the path steering requirement is not hop-by-hop explicit path 1126 selection, but loose-hop selection. Routed adjacencies can also 1127 allow to operate BIER-TE across intermediate hop routers that do not 1128 support BIER-TE. 1130 ............... 1131 ...BFR1--... ...--L1-- BFR2... 1132 ... .Routers. ...--L2--/ 1133 ...BFR4--... ...------ BFR3... 1134 ............... | 1135 LO 1136 Network Area 1 1138 Figure 14: Routed Adjacencies Example 1140 Assume the requirement in the above picture is to explicitly steer 1141 traffic flows that have arrived at BFR1 or BFR4 via a shortest path 1142 in the routing underlay "Network Area 1" to one of the following 1143 three next segments: (1) BFR2 via link L1, (2) BFR2 via link L2, (3) 1144 via BFR3. 1146 To enable this, both BFR1 and BFR4 are set up with a forward_routed 1147 adjacency BitPosition towards an address of BFR2 on link L1, another 1148 forward_routed BitPosition towards an address of BFR2 on link L2 and 1149 a third forward_routed Bitposition towards a node address LO of BFR3. 1151 4.8.2. Supporting nodes without BIER-TE 1153 Routed adjacencies also enable incremental deployment of BIER-TE. 1154 Only the nodes through which BIER-TE traffic needs to be steered - 1155 with or without replication - need to support BIER-TE. Where they 1156 are not directly connected to each other, forward_routed adjacencies 1157 are used to pass over non BIER-TE enabled nodes. 1159 4.9. Reuse of BitPositions (without DNR) 1161 BitPositions can be re-used across multiple BFR to minimize the 1162 number of BP needed. This happens when adjacencies on multiple BFR 1163 use the DNR flag as described above, but it can also be done for non- 1164 DNR adjacencies. This section only discussses this non-DNR case. 1166 Because BP are reset after passing a BFR with an adjacency for that 1167 BP, reuse of BP across multiple BFR does not introduce any problems 1168 with duplicates or loops that do not also exist when every adjacency 1169 has a unique BP: Instead of setting one BP in a BitString that is 1170 reused in N-adjacencies, one would get the same or worse results if 1171 each of these adjacencies had a unique BP and all of them where set 1172 in the BitString. Instead, based on the case, BPs can be reused 1173 without limitation, or they introduce fewer path steering choices, or 1174 they do not work. 1176 BP cannot be reused across two BFR that would need to be passed 1177 sequentially for some path: The first BFR will reset the BP, so those 1178 paths cannot be built. BP can be set across BFR that would (A) only 1179 occur across different paths or (B) across different branches of the 1180 same tree. 1182 An example of (A) was given in Figure 13, where BP 0:7, BP 0:8 and BP 1183 0:9 are each reused across multiple BFR because a single packet/path 1184 would never be able to reach more than one BFR sharing the same BP. 1186 Assume the example was changed: BFR1 has no ECMP adjacency for BP 1187 0:6, but instead BP 0:5 with forward_connected to BFR2 and BP 0:6 1188 with forward_connected to BFR3. Packets with both BP 0:5 and BP 0:6 1189 would now be able to reach both BFR2 and BFR3 and the still existing 1190 re-use of BP 0:7 between BFR2 and BFR3 is a case of (B) where reuse 1191 of BP is perfect because it does not limit the set of useful path 1192 choices: 1194 If instead of reusing BP 0:7, BFR3 used a separate BP 0:10 for its 1195 ECMP adjacency, no useful additional path steering options would be 1196 enabled. If duplicates at BFR10 where undesirable, this would be 1197 done by not setting BP 0:5 and BP 0:6 for the same packet. If the 1198 duplicates where desirable (e.g.: resilient transmission), the 1199 additional BP 0:10 would also not render additional value. 1201 Reuse may also save BPs in larger topologies. Consider the topology 1202 shown in Figure 17, but only the following explanations: A BFIR/ 1203 sender (e.g.: video headend) is attached to area 1, and area 2...6 1204 contain receivers/BFER. Assume each area had a distribution ring, 1205 each with two BPs to indicate the direction (as explained in before). 1206 These two BPs could be reused across the 5 areas. Packets would be 1207 replicated through other BPs to the desired subset of areas, and once 1208 a packet copy reaches the ring of the area, the two ring BPs come 1209 into play. This reuse is a case of (B), but it limits the topology 1210 choices: Packets can only flow around the same direction in the rings 1211 of all areas. This may or may not be acceptable based on the desired 1212 path steering options: If resilient transmission is the path 1213 engineering goal, then it is likely a good optimization, if the 1214 bandwidth of each ring was to be optimized separately, it would not 1215 be a good limitation. 1217 4.10. Summary of BP optimizations 1219 This section reviewed a range of techniques by which a BIER-TE 1220 Controller can create a BIER-TE topology in a way that minimizes the 1221 number of necessary BPs. 1223 Without any optimization, a BIER-TE Controller would attempt to map 1224 the network subnet topology 1:1 into the BIER-TE topology and every 1225 subnet adjacent neighbor requires a forward_connected BP and every 1226 BFER requires a local_decap BP. 1228 The optimizations described are then as follows: 1230 o P2p links require only one BP (Section 4.1). 1232 o All leaf-BFER can share a single local_decap BP (Section 4.3). 1234 o A LAN with N BFR needs at most N BP (one for each BFR). It only 1235 needs one BP for all those BFR tha are not redundanty connected to 1236 multiple LANs (Section 4.4). 1238 o A hub with p2p connections to multiple non-leaf-BFER spokes can 1239 share one BP to all spokes if traffic can be flooded to all 1240 spokes, e.g.: because of no bandwidth concerns or dense receiver 1241 sets (Section 4.5). 1243 o Rings of BFR can be built with just two BP (one for each 1244 direction) except for BFR with multiple ring connections - similar 1245 to LANs (Section 4.6). 1247 o ECMP adjacencies to N neighbors can replace N BP with 1 BP. 1248 Multihop ECMP can avoid polarization through different seeds of 1249 the ECMP algorithm (Section 4.7). 1251 o Routed adjacencies allow to "tunnel" across non-BIER-TE capable 1252 routers and across BIER-TE capable routers where no traffic- 1253 steering or replications are required (Section 4.8). 1255 o BP can generally be reused across nodes that do not need to be 1256 consecutive in paths, but depending on scenario, this may limit 1257 the feasible path steering options (Section 4.9). 1259 Note that the described list of optimizations is not exhaustive. 1260 Especially when the set of required path steering choices is limited 1261 and the set of possible subsets of BFER that should be able to 1262 receive traffic is limited, further optimizations of BP are possible. 1263 The hub & spoke optimization is a simple example of such traffic 1264 pattern dependent optimizations. 1266 5. Avoiding loops and duplicates 1268 5.1. Loops 1270 Whenever BIER-TE creates a copy of a packet, the BitString of that 1271 copy will have all BitPositions cleared that are associated with 1272 adjacencies on the BFR. This inhibits looping of packets. The only 1273 exception are adjacencies with DNR set. 1275 With DNR set, looping can happen. Consider in the ring picture that 1276 link L4 from BFR3 is plugged into the L1 interface of BFRa. This 1277 creates a loop where the rings clockwise BitPosition is never reset 1278 for copies of the packets traveling clockwise around the ring. 1280 To inhibit looping in the face of such physical misconfiguration, 1281 only forward_connected adjacencies are permitted to have DNR set, and 1282 the link layer port unique unicast destination address of the 1283 adjacency (e.g. MAC address) protects against closing the loop. 1284 Link layers without port unique link layer addresses should not be 1285 used with the DNR flag set. 1287 5.2. Duplicates 1289 Duplicates happen when the topology of the BitString is not a tree 1290 but redundantly connecting BFRs with each other. The BIER-TE 1291 Controller must therefore ensure to only create BitStrings that are 1292 trees in the topology. 1294 When links are incorrectly physically re-connected before the BIER-TE 1295 Controller updates BitStrings in BFIRs, duplicates can happen. Like 1296 loops, these can be inhibited by link layer addressing in 1297 forward_connected adjacencies. 1299 If interface or loopback addresses used in forward_routed adjacencies 1300 are moved from one BFR to another, duplicates can equally happen. 1301 Such re-addressing operations must be coordinated with the BIER-TE 1302 Controller. 1304 6. BIER-TE Forwarding Pseudocode 1306 The following simplified pseudocode for BIER-TE forwarding is using 1307 BIER forwarding pseudocode of [RFC8279], section 6.5 with the one 1308 modification necessary to support basic BIER-TE forwarding. Like the 1309 BIER pseudo forwarding code, for simplicity it does hide the details 1310 of the adjacency processing inside PacketSend() which can be 1311 forward_connected, forward_routed or local_decap. 1313 void ForwardBitMaskPacket_withTE (Packet) 1314 { 1315 SI=GetPacketSI(Packet); 1316 Offset=SI*BitStringLength; 1317 for (Index = GetFirstBitPosition(Packet->BitString); Index ; 1318 Index = GetNextBitPosition(Packet->BitString, Index)) { 1319 F-BM = BIFT[Index+Offset]->F-BM; 1320 if (!F-BM) continue; 1321 BFR-NBR = BIFT[Index+Offset]->BFR-NBR; 1322 PacketCopy = Copy(Packet); 1323 PacketCopy->BitString &= F-BM; [2] 1324 PacketSend(PacketCopy, BFR-NBR); 1325 // The following must not be done for BIER-TE: 1326 // Packet->BitString &= ~F-BM; [1] 1327 } 1328 } 1330 Figure 15: Simplified BIER-TE Forwarding Pseudocode 1332 The difference is that in BIER-TE, step [1] must not be performed, 1333 but is replaced with [2] (when the forwarding plane algorithm is 1334 implemented verbatim as shown above). 1336 In BIER, the F-BM of a BP has all BP set that are meant to be 1337 forwarded via the same neighbor. It is used to reset those BP in the 1338 packet after the first copy to this neighbor has been made to inhibit 1339 multiple copies to the same neighbor. 1341 In BIER-TE, the F-BM of a particular BP with an adjacency is the list 1342 of all BPs with an adjacency on this BFR except the particular BP 1343 itself if it has an adjacency with the DNR bit set. The F-BM is used 1344 to reset the F-BM BPs before creating copies. 1346 In BIER, the order of BPs impacts the result of forwarding because of 1347 [1]. In BIER-TE, forwarding is not impacted by the order of BPs. It 1348 is therefore possible to further optimize forwarding than in BIER. 1349 For example, BIER-TE forwarding can be parallelized such that a 1350 parallel instance (such as an egres linecard) can process any subset 1351 of BPs without any considerations for the other BPs - and without any 1352 prior, cross-BP shared processing. 1354 The above simplified pseudocode is elaborated further as follows: 1356 o This pseudocode eliminates per-bit F-BM, therefore reducing state 1357 by BitStringLength^2*SI and eliminating the need for per-packet- 1358 copy masking operation except for adjacencies with DNR flag set: 1360 * AdjacentBits[SI] are bits with a non-empty list of adjacencies. 1361 This can be computed whenever the BIER-TE Controller updates 1362 the adjacencies. 1364 * Only the AdjacentBits need to be examined in the loop for 1365 packet copies. 1367 * The packets BitString is masked with those AdjacentBits on 1368 ingress to avoid packets looping. 1370 o The code loops over the adjacencies because there may be more than 1371 one adjacency for a bit. 1373 o When an adjacency has the DNR bit, the bit is set in the packet 1374 copy (to save bits in rings for example). 1376 o The ECMP adjacency is shown. Its parameters are a 1377 ListOfAdjacencies from which one is picked. 1379 o The forward_local, forward_routed, local_decap adjacencies are 1380 shown with their parameters. 1382 void ForwardBitMaskPacket_withTE (Packet) 1383 { 1384 SI=GetPacketSI(Packet); 1385 Offset=SI*BitStringLength; 1386 AdjacentBitstring = Packet->BitString &= ~AdjacentBits[SI]; 1387 Packet->BitString &= AdjacentBits[SI]; 1388 for (Index = GetFirstBitPosition(AdjacentBits); Index ; 1389 Index = GetNextBitPosition(AdjacentBits, Index)) { 1390 foreach adjacency BIFT[Index+Offset] { 1391 if(adjacency == ECMP(ListOfAdjacencies, seed) ) { 1392 I = ECMP_hash(sizeof(ListOfAdjacencies), 1393 Packet->Entropy, seed); 1394 adjacency = ListOfAdjacencies[I]; 1395 } 1396 PacketCopy = Copy(Packet); 1397 switch(adjacency) { 1398 case forward_connected(interface,neighbor,DNR): 1399 if(DNR) 1400 PacketCopy->BitString |= 2<<(Index-1); 1401 SendToL2Unicast(PacketCopy,interface,neighbor); 1403 case forward_routed({VRF},neighbor): 1404 SendToL3(PacketCopy,{VRF,}l3-neighbor); 1406 case local_decap({VRF},neighbor): 1407 DecapBierHeader(PacketCopy); 1408 PassTo(PacketCopy,{VRF,}Packet->NextProto); 1409 } 1410 } 1411 } 1412 } 1414 Figure 16: BIER-TE Forwarding Pseudocode 1416 7. Managing SI, subdomains and BFR-ids 1418 When the number of bits required to represent the necessary hops in 1419 the topology and BFER exceeds the supported bitstring length, 1420 multiple SI and/or subdomains must be used. This section discusses 1421 how. 1423 BIER-TE forwarding does not require the concept of BFR-id, but 1424 routing underlay, flow overlay and BIER headers may. This section 1425 also discusses how BFR-ids can be assigned to BFIR/BFER for BIER-TE. 1427 7.1. Why SI and sub-domains 1429 For BIER and BIER-TE forwarding, the most important result of using 1430 multiple SI and/or subdomains is the same: Packets that need to be 1431 sent to BFER in different SI or subdomains require different BIER 1432 packets: each one with a bitstring for a different (SI,subdomain) 1433 combination. Each such bitstring uses one bitstring length sized SI 1434 block in the BIFT of the subdomain. We call this a BIFT:SI (block). 1436 For BIER and BIER-TE forwarding itself there is also no difference 1437 whether different SI and/or sub-domains are chosen, but SI and 1438 subdomain have different purposes in the BIER architecture shared by 1439 BIER-TE. This impacts how operators are managing them and how 1440 especially flow overlays will likely use them. 1442 By default, every possible BFIR/BFER in a BIER network would likely 1443 be given a BFR-id in subdomain 0 (unless there are > 64k BFIR/BFER). 1445 If there are different flow services (or service instances) requiring 1446 replication to different subsets of BFER, then it will likely not be 1447 possible to achieve the best replication efficiency for all of these 1448 service instances via subdomain 0. Ideal replication efficiency for 1449 N BFER exists in a subdomain if they are split over not more than 1450 ceiling(N/bitstring-length) SI. 1452 If service instances justify additional BIER:SI state in the network, 1453 additional subdomains will be used: BFIR/BFER are assigned BFIR-id in 1454 those subdomains and each service instance is configured to use the 1455 most appropriate subdomain. This results in improved replication 1456 efficiency for different services. 1458 Even if creation of subdomains and assignment of BFR-id to BFIR/BFER 1459 in those subdomains is automated, it is not expected that individual 1460 service instances can deal with BFER in different subdomains. A 1461 service instance may only support configuration of a single subdomain 1462 it should rely on. 1464 To be able to easily reuse (and modify as little as possible) 1465 existing BIER procedures including flow-overlay and routing underlay, 1466 when BIER-TE forwarding is added, we therefore reuse SI and subdomain 1467 logically in the same way as they are used in BIER: All necessary 1468 BFIR/BFER for a service use a single BIER-TE BIFT and are split 1469 across as many SI as necessary (see below). Different services may 1470 use different subdomains that primarily exist to provide more 1471 efficient replication (and for BIER-TE desirable path steering) for 1472 different subsets of BFIR/BFER. 1474 7.2. Bit assignment comparison BIER and BIER-TE 1476 In BIER, bitstrings only need to carry bits for BFER, which leads to 1477 the model that BFR-ids map 1:1 to each bit in a bitstring. 1479 In BIER-TE, bitstrings need to carry bits to indicate not only the 1480 receiving BFER but also the intermediate hops/links across which the 1481 packet must be sent. The maximum number of BFER that can be 1482 supported in a single bitstring or BIFT:SI depends on the number of 1483 bits necessary to represent the desired topology between them. 1485 "Desired" topology because it depends on the physical topology, and 1486 on the desire of the operator to allow for explicit path steeering 1487 across every single hop (which requires more bits), or reducing the 1488 number of required bits by exploiting optimizations such as unicast 1489 (forward_route), ECMP or flood (DNR) over "uninteresting" sub-parts 1490 of the topology - e.g. parts where different trees do not need to 1491 take different paths due to path steering reasons. 1493 The total number of bits to describe the topology vs. the BFER in a 1494 BIFT:SI can range widely based on the size of the topology and the 1495 amount of alternative paths in it. The higher the percentage, the 1496 higher the likelihood, that those topology bits are not just BIER-TE 1497 overhead without additional benefit, but instead that they will allow 1498 to express desirable path steering alternatives. 1500 7.3. Using BFR-id with BIER-TE 1502 Because there is no 1:1 mapping between bits in the bitstring and 1503 BFER, BIER-TE cannot simply rely on the BIER 1:1 mapping between bits 1504 in a bitstring and BFR-id. 1506 In BIER, automatic schemes could assign all possible BFR-ids 1507 sequentially to BFERs. This will not work in BIER-TE. In BIER-TE, 1508 the operator or BIER-TE Controller has to determine a BFR-id for each 1509 BFER in each required subdomain. The BFR-id may or may not have a 1510 relationship with a bit in the bitstring. Suggestions are detailed 1511 below. Once determined, the BFR-id can then be configured on the 1512 BFER and used by flow overlay, routing underlay and the BIER header 1513 almost the same as the BFR-id in BIER. 1515 The one exception are application/flow-overlays that automatically 1516 calculate the bitstring(s) of BIER packets by converting BFR-id to 1517 bits. In BIER-TE, this operation can be done in two ways: 1519 "Independent branches": For a given application or (set of) trees, 1520 the branches from a BFIR to every BFER are independent of the 1521 branches to any other BFER. For example, shortest part trees have 1522 independent branches. 1524 "Interdependent branches": When a BFER is added or deleted from a 1525 particular distribution tree, branches to other BFER still in the 1526 tree may need to change. Steiner tree are examples of dependent 1527 branch trees. 1529 If "independent branches" are sufficient, the BIER-TE Controller can 1530 provide to such applications for every BFR-id a SI:bitstring with the 1531 BIER-TE bits for the branch towards that BFER. The application can 1532 then independently calculate the SI:bitstring for all desired BFER by 1533 OR'ing their bitstrings. 1535 If "interdependent branches" are required, the application could call 1536 a BIER-TE Controller API with the list of required BFER-id and get 1537 the required bitstring back. Whenever the set of BFER-id changes, 1538 this is repeated. 1540 Note that in either case (unlike in BIER), the bits in BIER-TE may 1541 need to change upon link/node failure/recovery, network expansion and 1542 network resource consumption by other traffic as part of traffic 1543 engineering goals (e.g.: re-optimization of lower priority traffic 1544 flows). Interactions between such BFIR applications and the BIER-TE 1545 Controller do therefore need to support dynamic updates to the 1546 bitstrings. 1548 7.4. Assigning BFR-ids for BIER-TE 1550 For a non-leaf BFER, there is usually a single bit k for that BFER 1551 with a local_decap() adjacency on the BFER. The BFR-id for such a 1552 BFER is therefore most easily the one it would have in BIER: SI * 1553 bitstring-length + k. 1555 As explained earlier in the document, leaf BFERs do not need such a 1556 separate bit because the fact alone that the BIER-TE packet is 1557 forwarded to the leaf BFER indicates that the BFER should decapsulate 1558 it. Such a BFER will have one or more bits for the links leading 1559 only to it. The BFR-id could therefore most easily be the BFR-id 1560 derived from the lowest bit for those links. 1562 These two rules are only recommendations for the operator or BIER-TE 1563 Controller assigning the BFR-ids. Any allocation scheme can be used, 1564 the BFR-ids just need to be unique across BFRs in each subdomain. 1566 It is not currently determined if a single subdomain could or should 1567 be allowed to forward both BIER and BIER-TE packets. If this should 1568 be supported, there are two options: 1570 A. BIER and BIER-TE have different BFR-id in the same subdomain. 1571 This allows higher replication efficiency for BIER because their BFR- 1572 id can be assigned sequentially, while the bitstrings for BIER-TE 1573 will have also the additional bits for the topology. There is no 1574 relationship between a BFR BIER BFR-id and BIER-TE BFR-id. 1576 B. BIER and BIER-TE share the same BFR-id. The BFR-id are assigned 1577 as explained above for BIER-TE and simply reused for BIER. The 1578 replication efficiency for BIER will be as low as that for BIER-TE in 1579 this approach. Depending on topology, only the same 20%..80% of bits 1580 as possible for BIER-TE can be used for BIER. 1582 7.5. Example bit allocations 1584 7.5.1. With BIER 1586 Consider a network setup with a bitstring length of 256 for a network 1587 topology as shown in the picture below. The network has 6 areas, 1588 each with ca. 170 BFR, connecting via a core with some larger (core) 1589 BFR. To address all BFER with BIER, 4 SI are required. To send a 1590 BIER packet to all BFER in the network, 4 copies need to be sent by 1591 the BFIR. On the BFIR it does not make a difference how the BFR-id 1592 are allocated to BFER in the network, but for efficiency further down 1593 in the network it does make a difference. 1595 area1 area2 area3 1596 BFR1a BFR1b BFR2a BFR2b BFR3a BFR3b 1597 | \ / \ / | 1598 ................................ 1599 . Core . 1600 ................................ 1601 | / \ / \ | 1602 BFR4a BFR4b BFR5a BFR5b BFR6a BFR6b 1603 area4 area5 area6 1605 Figure 17: Scaling BIER-TE bits by reuse 1607 With random allocation of BFR-id to BFER, each receiving area would 1608 (most likely) have to receive all 4 copies of the BIER packet because 1609 there would be BFR-id for each of the 4 SI in each of the areas. 1610 Only further towards each BFER would this duplication subside - when 1611 each of the 4 trees runs out of branches. 1613 If BFR-id are allocated intelligently, then all the BFER in an area 1614 would be given BFR-id with as few as possible different SI. Each 1615 area would only have to forward one or two packets instead of 4. 1617 Given how networks can grow over time, replication efficiency in an 1618 area will also easily go down over time when BFR-id are network wide 1619 allocated sequentially over time. An area that initially only has 1620 BFR-id in one SI might end up with many SI over a longer period of 1621 growth. Allocating SIs to areas with initially sufficiently many 1622 spare bits for growths can help to alleviate this issue. Or renumber 1623 BFR-id after network expansion. In this example one may consider to 1624 use 6 SI and assign one to each area. 1626 This example shows that intelligent BFR-id allocation within at least 1627 subdomain 0 can even be helpful or even necessary in BIER. 1629 7.5.2. With BIER-TE 1631 In BIER-TE one needs to determine a subset of the physical topology 1632 and attached BFER so that the "desired" representation of this 1633 topology and the BFER fit into a single bitstring. This process 1634 needs to be repeated until the whole topology is covered. 1636 Once bits/SIs are assigned to topology and BFER, BFR-id is just a 1637 derived set of identifiers from the operator/BIER-TE Controller as 1638 explained above. 1640 Every time that different sub-topologies have overlap, bits need to 1641 be repeated across the bitstrings, increasing the overall amount of 1642 bits required across all bitstring/SIs. In the worst case, random 1643 subsets of BFER are assigned to different SI. This is much worse 1644 than in BIER because it not only reduces replication efficiency with 1645 the same number of overall bits, but even further - because more bits 1646 are required due to duplication of bits for topology across multiple 1647 SI. Intelligent BFER to SI assignment and selecting specific 1648 "desired" subtopologies can minimize this problem. 1650 To set up BIER-TE efficiently for above topology, the following bit 1651 allocation methods can be used. This method can easily be expanded 1652 to other, similarly structured larger topologies. 1654 Each area is allocated one or more SI depending on the number of 1655 future expected BFER and number of bits required for the topology in 1656 the area. In this example, 6 SI, one per area. 1658 In addition, we use 4 bits in each SI: bia, bib, bea, beb: bit 1659 ingress a, bit ingress b, bit egress a, bit egress b. These bits 1660 will be used to pass BIER packets from any BFIR via any combination 1661 of ingress area a/b BFR and egress area a/b BFR into a specific 1662 target area. These bits are then set up with the right 1663 forward_routed adjacencies on the BFIR and area edge BFR: 1665 On all BFIR in an area j, bia in each BIFT:SI is populated with the 1666 same forward_routed(BFRja), and bib with forward_routed(BFRjb). On 1667 all area edge BFR, bea in BIFT:SI=k is populated with 1668 forward_routed(BFRka) and beb in BIFT:SI=k with 1669 forward_routed(BFRkb). 1671 For BIER-TE forwarding of a packet to some subset of BFER across all 1672 areas, a BFIR would create at most 6 copies, with SI=1...SI=6, In 1673 each packet, the bits indicate bits for topology and BFER in that 1674 topology plus the four bits to indicate whether to pass this packet 1675 via the ingress area a or b border BFR and the egress area a or b 1676 border BFR, therefore allowing path steering for those two "unicast" 1677 legs: 1) BFIR to ingress are edge and 2) core to egress area edge. 1678 Replication only happens inside the egress areas. For BFER in the 1679 same area as in the BFIR, these four bits are not used. 1681 7.6. Summary 1683 BIER-TE can like BIER support multiple SI within a sub-domain to 1684 allow re-using the concept of BFR-id and therefore minimize BIER-TE 1685 specific functions in underlay routing, flow overlay methods and BIER 1686 headers. 1688 The number of BFIR/BFER possible in a subdomain is smaller than in 1689 BIER because BIER-TE uses additional bits for topology. 1691 Subdomains can in BIER-TE be used like in BIER to create more 1692 efficient replication to known subsets of BFER. 1694 Assigning bits for BFER intelligently into the right SI is more 1695 important in BIER-TE than in BIER because of replication efficiency 1696 and overall amount of bits required. 1698 8. BIER-TE and Segment Routing 1700 SR aims to enable lightweight path steering via loose source routing. 1701 Compared to its more heavy-weight predecessor RSVP-TE, SR does for 1702 example not require per-path signaling to each of these hops. 1704 BIER-TE supports the same design philosophy for multicast. Like in 1705 SR, it relies on source-routing - via the definition of a BitString. 1706 Like SR, it only requires to consider the "hops" on which either 1707 replication has to happen, or across which the traffic should be 1708 steered (even without replication). Any other hops can be skipped 1709 via the use of routed adjacencies. 1711 BIER-TE BitPosition (BP) can be understood as the BIER-TE equivalent 1712 of "forwarding segments" in SR, but they have a different scope than 1713 SR forwarding segments. Whereas forwarding segments in SR are global 1714 or local, BPs in BIER-TE have a scope that is the group of BFR(s) 1715 that have adjacencies for this BP in their BIFT. This can be called 1716 "adjacency" scoped forwarding segments. 1718 Adjacency scope could be global, but then every BFR would need an 1719 adjacency for this BP, for example a forward_routed adjacency with 1720 encapsulation to the global SR SID of the destination. Such a BP 1721 would always result in ingress replication though. The first BFR 1722 encountering this BP would directly replicate to it. Only by using 1723 non-global adjacency scope for BPs can traffic be steered and 1724 replicated on non-ingress BFR. 1726 SR can naturally be combined with BIER-TE and help to optimize it. 1727 For example, instead of defining BitPositions for non-replicating 1728 hops, it is equally possible to use segment routing encapsulations 1729 (eg: MPLS label stacks) for the encapsulation of "forward_routed" 1730 adjacencies. 1732 Note that BIER itself can also be seen to be similar to SR. BIER BPs 1733 act as global destination Node-SIDs and the BIER bitstring is simply 1734 a highly optimized mechanism to indicate multiple such SIDS and let 1735 the network take care of effectively replicating the packet hop-by- 1736 hop to each destination Node-SID. What BIER does not allow is to 1737 indicate intermediate hops, or terms of SR the ability to indicate a 1738 sequence of SID to reach the destination. This is what BIER-TE and 1739 its adjacency scoped BP enables. 1741 Both BIER and BIER-TE allow BFIR to "opportunistically" copy packets 1742 to a set of desired BFER on a packet-by-packet basis. In BIER, this 1743 is done by OR'ing the BP for the desired BFER. In BIER-TE this can 1744 be done by OR'ing for each desired BFER a bitstring using the 1745 "independent branches" approach described in Section 7.3 and 1746 therefore also indicating the engineered path towards each desired 1747 BFER. This is the approach that 1748 [I-D.ietf-bier-multicast-http-response] relies on. 1750 9. Security Considerations 1752 The security considerations are the same as for BIER with the 1753 following differences: 1755 BFR-ids and BFR-prefixes are not used in BIER-TE, nor are procedures 1756 for their distribution, so these are not attack vectors against BIER- 1757 TE. 1759 10. IANA Considerations 1761 This document requests no action by IANA. 1763 11. Acknowledgements 1765 The authors would like to thank Greg Shepherd, Ijsbrand Wijnands, 1766 Neale Ranns, Dirk Trossen, Sandy Zheng, Lou Berger and Jeffrey Zhang 1767 for their reviews and suggestions. 1769 12. Change log [RFC Editor: Please remove] 1771 draft-ietf-bier-te-arch: 1773 08: Incorporated (with hopefully acceptable fixes) for Lou 1774 suggested section 2.5, TE considerations. 1776 Fixes are primarily to the point to a) emphasize that BIER-TE does 1777 not depend on the routing underlay unless forward_routed 1778 adjacencies are used, and b) that the allocation and tracking of 1779 resources does not explicitly have to be tied to BPs, because they 1780 are just steering labels. Instead, it would ideally come from 1781 per-hop resource management that can be maintained only via local 1782 accounting in the controller. 1784 07: Further reworking text for Lou. 1786 Renamed BIER-PE to BIER-TE standing for "Tree Engineering" after 1787 votes from BIER WG. 1789 Removed section 1.1 (introduced by version 06) because not 1790 considered necessary in this doc by Lou (for framework doc). 1792 Added [RFC editor pls. remove] Section to explain name change to 1793 future reviewers. 1795 06: Concern by Lou Berger re. BIER-TE as full traffic engineering 1796 solution. 1798 Changed title "Traffic Engineering" to "Path Engineering" 1800 Added intro section of relationship BIER-PE to traffic 1801 engineering. 1803 Changed "traffic engineering" term in text" to "path engineering", 1804 where appropriate 1806 Other: 1808 Shortened "BIER-TE Controller Host" to "BIER-TE Controller". 1809 Fixed up all instances of controller to do this. 1811 05: Review Jeffrey Zhang. 1813 Part 2: 1815 4.3 added note about leaf-BFER being also a propery of routing 1816 setup. 1818 4.7 Added missing details from example to avoid confusion with 1819 routed adjacencies, also compressed explanatory text and better 1820 justification why seed is explicitly configured by controller. 1822 4.9 added section discussing generic reuse of BP methods. 1824 4.10 added section summarizing BP optimizations of section 4. 1826 6. Rewrote/compressed explanation of comparison BIER/BIER-TE 1827 forwarding difference. Explained benefit of BIER-TE per-BP 1828 forwarding being independent of forwarding for other BPs. 1830 Part 1: 1832 Explicitly ue forwarded_connected adjcency in ECMP adjcency 1833 examples to avoid confusion. 1835 4.3 Add picture as example for leav vs. non-leaf BFR in topology. 1836 Improved description. 1838 4.5 Exampe for traffic that can be broadcast -> for single BP in 1839 hub&spoke. 1841 4.8.1 Simplified example picture for routed adjacency, explanatory 1842 text. 1844 Review from Dirk Trossen: 1846 Fixed up explanation of ICC paper vs. bloom filter. 1848 04: spell check run. 1850 Addded remaining fixes for Sandys (Zhang Zheng) review: 1852 4.7 Enhance ECMP explanations: 1854 example ECMP algorithm, highlight that doc does not standardize 1855 ECMP algorithm. 1857 Review from Dirk Trossen: 1859 1. Added mentioning of prior work for traffic engineered paths 1860 with bloom filters. 1862 2. Changed title from layers to components and added "BIER-TE 1863 control plane" to "BIER-TE Controller" to make it clearer, what it 1864 does. 1866 2.2.3. Added reference to I-D.ietf-bier-multicast-http-response 1867 as an example solution. 1869 2.3. clarified sentence about resetting BPs before sending copies 1870 (also forgot to mention DNR here). 1872 3.4. Added text saying this section will be removed unless IESG 1873 review finds enough redeeming value in this example given how -03 1874 introduced section 1.1 with basic examples. 1876 7.2. Removed explicit numbers 20%/80% for number of topology bits 1877 in BIER-TE, replaced with more vague (high/low) description, 1878 because we do not have good reference material Added text saying 1879 this section will be removed unless IESG review finds enough 1880 redeeming value in this example given how -03 introduced section 1881 1.1 with basic examples. 1883 many typos fixed. Thanks a lot. 1885 03: Last call textual changes by authors to improve readability: 1887 removed Wolfgang Braun as co-authors (as requested). 1889 Improved abstract to be more explanatory. Removed mentioning of 1890 FRR (not concluded on so far). 1892 Added new text into Introduction section because the text was too 1893 difficult to jump into (too many forward pointers). This 1894 primarily consists of examples and the early introduction of the 1895 BIER-TE Topology concept enabled by these examples. 1897 Amended comparison to SR. 1899 Changed syntax from [VRF] to {VRF} to indicate its optional and to 1900 make idnits happy. 1902 Split references into normative / informative, added references. 1904 02: Refresh after IETF104 discussion: changed intended status back 1905 to standard. Reasoning: 1907 Tighter review of standards document == ensures arch will be 1908 better prepared for possible adoption by other WGs (e.g. DetNet) 1909 or std. bodies. 1911 Requirement against the degree of existing implementations is self 1912 defined by the WG. BIER WG seems to think it is not necessary to 1913 apply multiple interoperating implementations against an 1914 architecture level document at this time to make it qualify to go 1915 to standards track. Also, the levels of support introduced in -01 1916 rev. should allow all BIER forwarding engines to also be able to 1917 support the base level BIER-TE forwarding. 1919 01: Added note comparing BIER and SR to also hopefully clarify 1920 BIER-TE vs. BIER comparison re. SR. 1922 - added requirements section mandating only most basic BIER-TE 1923 forwarding features as MUST. 1925 - reworked comparison with BIER forwarding section to only 1926 summarize and point to pseudocode section. 1928 - reworked pseudocode section to have one pseudocode that mirrors 1929 the BIER forwarding pseudocode to make comparison easier and a 1930 second pseudocode that shows the complete set of BIER-TE 1931 forwarding options and simplification/optimization possible vs. 1932 BIER forwarding. Removed MyBitsOfInterest (was pure 1933 optimization). 1935 - Added captions to pictures. 1937 - Part of review feedback from Sandy (Zhang Zheng) integrated. 1939 00: Changed target state to experimental (WG conclusion), updated 1940 references, mod auth association. 1942 - Source now on http://www.github.com/toerless/bier-te-arch 1944 - Please open issues on the github for change/improvement requests 1945 to the document - in addition to posting them on the list 1946 (bier@ietf.). Thanks!. 1948 draft-eckert-bier-te-arch: 1950 06: Added overview of forwarding differences between BIER, BIER- 1951 TE. 1953 05: Author affiliation change only. 1955 04: Added comparison to Live-Live and BFIR to FRR section 1956 (Eckert). 1958 04: Removed FRR content into the new FRR draft [I-D.eckert-bier- 1959 te-frr] (Braun). 1961 - Linked FRR information to new draft in Overview/Introduction 1963 - Removed BTAFT/FRR from "Changes in the network topology" 1965 - Linked new draft in "Link/Node Failures and Recovery" 1967 - Removed FRR from "The BIER-TE Forwarding Layer" 1969 - Moved FRR section to new draft 1971 - Moved FRR parts of Pseudocode into new draft 1973 - Left only non FRR parts 1975 - removed FrrUpDown(..) and //FRR operations in 1976 ForwardBierTePacket(..) 1978 - New draft contains FrrUpDown(..) and ForwardBierTePacket(Packet) 1979 from bier-arch-03 1981 - Moved "BIER-TE and existing FRR to new draft 1983 - Moved "BIER-TE and Segment Routing" section one level up 1985 - Thus, removed "Further considerations" that only contained this 1986 section 1988 - Added Changes for version 04 1990 03: Updated the FRR section. Added examples for FRR key concepts. 1991 Added BIER-in-BIER tunneling as option for tunnels in backup 1992 paths. BIFT structure is expanded and contains an additional 1993 match field to support full node protection with BIER-TE FRR. 1995 03: Updated FRR section. Explanation how BIER-in-BIER 1996 encapsulation provides P2MP protection for node failures even 1997 though the routing underlay does not provide P2MP. 1999 02: Changed the definition of BIFT to be more inline with BIER. 2000 In revs. up to -01, the idea was that a BIFT has only entries for 2001 a single bitstring, and every SI and subdomain would be a separate 2002 BIFT. In BIER, each BIFT covers all SI. This is now also how we 2003 define it in BIER-TE. 2005 02: Added Section 7 to explain the use of SI, subdomains and BFR- 2006 id in BIER-TE and to give an example how to efficiently assign 2007 bits for a large topology requiring multiple SI. 2009 02: Added further detailed for rings - how to support input from 2010 all ring nodes. 2012 01: Fixed BFIR -> BFER for section 4.3. 2014 01: Added explanation of SI, difference to BIER ECMP, 2015 consideration for Segment Routing, unicast FRR, considerations for 2016 encapsulation, explanations of BIER-TE Controller and CLI. 2018 00: Initial version. 2020 13. References 2022 13.1. Normative References 2024 [RFC8279] Wijnands, IJ., Ed., Rosen, E., Ed., Dolganow, A., 2025 Przygienda, T., and S. Aldrin, "Multicast Using Bit Index 2026 Explicit Replication (BIER)", RFC 8279, 2027 DOI 10.17487/RFC8279, November 2017, 2028 . 2030 [RFC8296] Wijnands, IJ., Ed., Rosen, E., Ed., Dolganow, A., 2031 Tantsura, J., Aldrin, S., and I. Meilik, "Encapsulation 2032 for Bit Index Explicit Replication (BIER) in MPLS and Non- 2033 MPLS Networks", RFC 8296, DOI 10.17487/RFC8296, January 2034 2018, . 2036 13.2. Informative References 2038 [I-D.dt-teas-rfc3272bis] 2039 Farrel, A., "Overview and Principles of Internet Traffic 2040 Engineering", draft-dt-teas-rfc3272bis-11 (work in 2041 progress), May 2020. 2043 [I-D.ietf-bier-multicast-http-response] 2044 Trossen, D., Rahman, A., Wang, C., and T. Eckert, 2045 "Applicability of BIER Multicast Overlay for Adaptive 2046 Streaming Services", draft-ietf-bier-multicast-http- 2047 response-03 (work in progress), February 2020. 2049 [I-D.ietf-roll-ccast] 2050 Bergmann, O., Bormann, C., Gerdes, S., and H. Chen, 2051 "Constrained-Cast: Source-Routed Multicast for RPL", 2052 draft-ietf-roll-ccast-01 (work in progress), October 2017. 2054 [I-D.qiang-detnet-large-scale-detnet] 2055 Qiang, L., Geng, X., Liu, B., Eckert, T., Geng, L., and G. 2056 Li, "Large-Scale Deterministic IP Network", draft-qiang- 2057 detnet-large-scale-detnet-05 (work in progress), September 2058 2019. 2060 [ICC] Reed, M., Al-Naday, M., Thomos, N., Trossen, D., 2061 Petropoulos, G., and S. Spirou, "Stateless multicast 2062 switching in software defined networks", IEEE 2063 International Conference on Communications (ICC), Kuala 2064 Lumpur, Malaysia, 2016, May 2016, 2065 . 2067 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 2068 Requirement Levels", BCP 14, RFC 2119, 2069 DOI 10.17487/RFC2119, March 1997, 2070 . 2072 Authors' Addresses 2074 Toerless Eckert (editor) 2075 Futurewei Technologies Inc. 2076 2330 Central Expy 2077 Santa Clara 95050 2078 USA 2080 Email: tte+ietf@cs.fau.de 2082 Gregory Cauchie 2083 Bouygues Telecom 2085 Email: GCAUCHIE@bouyguestelecom.fr 2086 Michael Menth 2087 University of Tuebingen 2089 Email: menth@uni-tuebingen.de