idnits 2.17.1 draft-ietf-bess-evpn-overlay-09.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 : ---------------------------------------------------------------------------- ** There is 1 instance of lines with control characters in the document. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (December 7, 2017) is 2303 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) == Missing Reference: 'RFC2119' is mentioned on line 166, but not defined == Unused Reference: 'KEYWORDS' is defined on line 1219, but no explicit reference was found in the text ** Obsolete normative reference: RFC 5512 (Obsoleted by RFC 9012) == Outdated reference: A later version (-10) exists of draft-ietf-bess-dci-evpn-overlay-04 == Outdated reference: A later version (-22) exists of draft-ietf-idr-tunnel-encaps-03 == Outdated reference: A later version (-13) exists of draft-ietf-nvo3-vxlan-gpe-03 Summary: 2 errors (**), 0 flaws (~~), 6 warnings (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 BESS Workgroup A. Sajassi (Editor) 3 INTERNET-DRAFT Cisco 4 Intended Status: Standards Track J. Drake (Editor) 5 Juniper 6 N. Bitar 7 Nokia 8 R. Shekhar 9 Juniper 10 J. Uttaro 11 AT&T 12 W. Henderickx 13 Nokia 15 Expires: May 7, 2018 December 7, 2017 17 A Network Virtualization Overlay Solution using EVPN 18 draft-ietf-bess-evpn-overlay-09 20 Abstract 22 This document specifies how Ethernet VPN (EVPN) can be used as a 23 Network Virtualization Overlay (NVO) solution and explores the 24 various tunnel encapsulation options over IP and their impact on the 25 EVPN control-plane and procedures. In particular, the following 26 encapsulation options are analyzed: VXLAN, NVGRE, and MPLS over GRE. 27 This document also specifies new multi-homing procedures for split- 28 horizon filtering and mass-withdraw. It also specifies EVPN route 29 constructions for VxLAN/NvGRE encapsulations and ASBR procedures for 30 multi-homing NV Edge devices. 32 Status of this Memo 34 This Internet-Draft is submitted to IETF in full conformance with the 35 provisions of BCP 78 and BCP 79. 37 Internet-Drafts are working documents of the Internet Engineering 38 Task Force (IETF), its areas, and its working groups. Note that 39 other groups may also distribute working documents as 40 Internet-Drafts. 42 Internet-Drafts are draft documents valid for a maximum of six months 43 and may be updated, replaced, or obsoleted by other documents at any 44 time. It is inappropriate to use Internet-Drafts as reference 45 material or to cite them other than as "work in progress." 47 The list of current Internet-Drafts can be accessed at 48 http://www.ietf.org/1id-abstracts.html 50 The list of Internet-Draft Shadow Directories can be accessed at 51 http://www.ietf.org/shadow.html 53 Copyright and License Notice 55 Copyright (c) 2017 IETF Trust and the persons identified as the 56 document authors. All rights reserved. 58 This document is subject to BCP 78 and the IETF Trust's Legal 59 Provisions Relating to IETF Documents 60 (http://trustee.ietf.org/license-info) in effect on the date of 61 publication of this document. Please review these documents 62 carefully, as they describe your rights and restrictions with respect 63 to this document. Code Components extracted from this document must 64 include Simplified BSD License text as described in Section 4.e of 65 the Trust Legal Provisions and are provided without warranty as 66 described in the Simplified BSD License. 68 Table of Contents 70 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4 71 2 Requirements Notation and Conventions . . . . . . . . . . . . . 5 72 3 Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . 5 73 4 EVPN Features . . . . . . . . . . . . . . . . . . . . . . . . . 6 74 5 Encapsulation Options for EVPN Overlays . . . . . . . . . . . . 7 75 5.1 VXLAN/NVGRE Encapsulation . . . . . . . . . . . . . . . . . 7 76 5.1.1 Virtual Identifiers Scope . . . . . . . . . . . . . . . 8 77 5.1.1.1 Data Center Interconnect with Gateway . . . . . . . 8 78 5.1.1.2 Data Center Interconnect without Gateway . . . . . . 9 79 5.1.2 Virtual Identifiers to EVI Mapping . . . . . . . . . . . 9 80 5.1.2.1 Auto Derivation of RT . . . . . . . . . . . . . . . 10 81 5.1.3 Constructing EVPN BGP Routes . . . . . . . . . . . . . 12 82 5.2 MPLS over GRE . . . . . . . . . . . . . . . . . . . . . . . 13 83 6 EVPN with Multiple Data Plane Encapsulations . . . . . . . . . 14 84 7 Single-Homing NVEs - NVE Residing in Hypervisor . . . . . . . . 15 85 7.1 Impact on EVPN BGP Routes & Attributes for VXLAN/NVGRE 86 Encapsulation . . . . . . . . . . . . . . . . . . . . . . . 15 87 7.2 Impact on EVPN Procedures for VXLAN/NVGRE Encapsulation . . 16 88 8 Multi-Homing NVEs - NVE Residing in ToR Switch . . . . . . . . 16 89 8.1 EVPN Multi-Homing Features . . . . . . . . . . . . . . . . 17 90 8.1.1 Multi-homed Ethernet Segment Auto-Discovery . . . . . . 17 91 8.1.2 Fast Convergence and Mass Withdraw . . . . . . . . . . . 17 92 8.1.3 Split-Horizon . . . . . . . . . . . . . . . . . . . . . 17 93 8.1.4 Aliasing and Backup-Path . . . . . . . . . . . . . . . . 17 94 8.1.5 DF Election . . . . . . . . . . . . . . . . . . . . . . 18 95 8.2 Impact on EVPN BGP Routes & Attributes . . . . . . . . . . . 19 96 8.3 Impact on EVPN Procedures . . . . . . . . . . . . . . . . . 19 97 8.3.1 Split Horizon . . . . . . . . . . . . . . . . . . . . . 19 98 8.3.2 Aliasing and Backup-Path . . . . . . . . . . . . . . . . 20 99 8.3.3 Unknown Unicast Traffic Designation . . . . . . . . . . 20 100 9 Support for Multicast . . . . . . . . . . . . . . . . . . . . . 21 101 10 Data Center Interconnections - DCI . . . . . . . . . . . . . . 22 102 10.1 DCI using GWs . . . . . . . . . . . . . . . . . . . . . . . 22 103 10.2 DCI using ASBRs . . . . . . . . . . . . . . . . . . . . . . 23 104 10.2.1 ASBR Functionality with Single-Homing NVEs . . . . . . 24 105 10.2.2 ASBR Functionality with Multi-Homing NVEs . . . . . . . 24 106 11 Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . 26 107 12 Security Considerations . . . . . . . . . . . . . . . . . . . 26 108 13 IANA Considerations . . . . . . . . . . . . . . . . . . . . . 27 109 14 References . . . . . . . . . . . . . . . . . . . . . . . . . . 27 110 14.1 Normative References . . . . . . . . . . . . . . . . . . . 27 111 14.2 Informative References . . . . . . . . . . . . . . . . . . 27 112 Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 113 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 29 115 1 Introduction 117 In the context of this document, a Network Virtualization Overlay 118 (NVO) is a solution to address the requirements of a multi-tenant 119 data center, especially one with virtualized hosts, e.g., Virtual 120 Machines (VMs) or virtual workloads. The key requirements of such a 121 solution, as described in [RFC7364], are: 123 - Isolation of network traffic per tenant 125 - Support for a large number of tenants (tens or hundreds of 126 thousands) 128 - Extending L2 connectivity among different VMs belonging to a given 129 tenant segment (subnet) across different PODs within a data center or 130 between different data centers 132 - Allowing a given VM to move between different physical points of 133 attachment within a given L2 segment 135 The underlay network for NVO solutions is assumed to provide IP 136 connectivity between NVO endpoints (NVEs). 138 This document describes how Ethernet VPN (EVPN) can be used as an NVO 139 solution and explores applicability of EVPN functions and procedures. 140 In particular, it describes the various tunnel encapsulation options 141 for EVPN over IP, and their impact on the EVPN control-plane and 142 procedures for two main scenarios: 144 a) single-homing NVEs - when a NVE resides in the hypervisor, and 145 b) multi-homing NVEs - when a NVE resides in a Top of Rack (ToR) 146 device 148 The possible encapsulation options for EVPN overlays that are 149 analyzed in this document are: 151 - VXLAN and NVGRE 152 - MPLS over GRE 154 Before getting into the description of the different encapsulation 155 options for EVPN over IP, it is important to highlight the EVPN 156 solution's main features, how those features are currently supported, 157 and any impact that the encapsulation has on those features. 159 2 Requirements Notation and Conventions 161 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 162 "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and 163 "OPTIONAL" in this document are to be interpreted as described in BCP 164 14 [RFC2119] [RFC8174] when, and only when, they appear in all 165 capitals, as shown here. 167 3 Terminology 169 Most of the terminology used in this documents comes from [RFC7432] 170 and [RFC7365]. 172 NVO: Network Virtualization Overlay 174 NVE: Network Virtualization Endpoint 176 VNI: Virtual Network Identifier (for VXLAN) 178 VSID: Virtual Subnet Identifier (for NVGRE) 180 EVPN: Ethernet VPN 182 EVI: An EVPN instance spanning the Provider Edge (PE) devices 183 participating in that EVPN. 185 MAC-VRF: A Virtual Routing and Forwarding table for Media Access 186 Control (MAC) addresses on a PE. 188 Ethernet Segment (ES): When a customer site (device or network) is 189 connected to one or more PEs via a set of Ethernet links, then that 190 set of links is referred to as an 'Ethernet segment'. 192 Ethernet Segment Identifier (ESI): A unique non-zero identifier that 193 identifies an Ethernet segment is called an 'Ethernet Segment 194 Identifier'. 196 Ethernet Tag: An Ethernet tag identifies a particular broadcast 197 domain, e.g., a VLAN. An EVPN instance consists of one or more 198 broadcast domains. 200 PE: Provider Edge device. 202 Single-Active Redundancy Mode: When only a single PE, among all the 203 PEs attached to an Ethernet segment, is allowed to forward traffic 204 to/from that Ethernet segment for a given VLAN, then the Ethernet 205 segment is defined to be operating in Single-Active redundancy mode. 207 All-Active Redundancy Mode: When all PEs attached to an Ethernet 208 segment are allowed to forward known unicast traffic to/from that 209 Ethernet segment for a given VLAN, then the Ethernet segment is 210 defined to be operating in All-Active redundancy mode. 212 4 EVPN Features 214 EVPN was originally designed to support the requirements detailed in 215 [RFC7209] and therefore has the following attributes which directly 216 address control plane scaling and ease of deployment issues. 218 1) Control plane information is distributed with BGP and Broadcast 219 and Multicast traffic is sent using a shared multicast tree or with 220 ingress replication. 222 2) Control plane learning is used for MAC (and IP) addresses instead 223 of data plane learning. The latter requires the flooding of unknown 224 unicast and ARP frames; whereas, the former does not require any 225 flooding. 227 3) Route Reflectors are used to reduce a full mesh of BGP sessions 228 among PE devices to a single BGP session between a PE and the RR. 229 Furthermore, RR hierarchy can be leveraged to scale the number of BGP 230 routes on the RR. 232 4) Auto-discovery via BGP is used to discover PE devices 233 participating in a given VPN, PE devices participating in a given 234 redundancy group, tunnel encapsulation types, multicast tunnel type, 235 multicast members, etc. 237 5) All-Active multihoming is used. This allows a given customer 238 device (CE) to have multiple links to multiple PEs, and traffic 239 to/from that CE fully utilizes all of these links. 241 6) When a link between a CE and a PE fails, the PEs for that EVI are 242 notified of the failure via the withdrawal of a single EVPN route. 243 This allows those PEs to remove the withdrawing PE as a next hop for 244 every MAC address associated with the failed link. This is termed 245 'mass withdrawal'. 247 7) BGP route filtering and constrained route distribution are 248 leveraged to ensure that the control plane traffic for a given EVI is 249 only distributed to the PEs in that EVI. 251 8) When a 802.1Q interface is used between a CE and a PE, each of the 252 VLAN ID (VID) on that interface can be mapped onto a bridge table 253 (for upto 4094 such bridge tables). All these bridge tables may be 254 mapped onto a single MAC-VRF (in case of VLAN-aware bundle service). 256 9) VM Mobility mechanisms ensure that all PEs in a given EVI know 257 the ES with which a given VM, as identified by its MAC and IP 258 addresses, is currently associated. 260 10) Route Targets are used to allow the operator (or customer) to 261 define a spectrum of logical network topologies including mesh, hub & 262 spoke, and extranets (e.g., a VPN whose sites are owned by different 263 enterprises), without the need for proprietary software or the aid of 264 other virtual or physical devices. 266 Because the design goal for NVO is millions of instances per common 267 physical infrastructure, the scaling properties of the control plane 268 for NVO are extremely important. EVPN and the extensions described 269 herein, are designed with this level of scalability in mind. 271 5 Encapsulation Options for EVPN Overlays 273 5.1 VXLAN/NVGRE Encapsulation 275 Both VXLAN and NVGRE are examples of technologies that provide a data 276 plane encapsulation which is used to transport a packet over the 277 common physical IP infrastructure between Network Virtualization 278 Edges (NVEs) - e.g., VXLAN Tunnel End Points (VTEPs) in VXLAN 279 network. Both of these technologies include the identifier of the 280 specific NVO instance, Virtual Network Identifier (VNI) in VXLAN and 281 Virtual Subnet Identifier (VSID) in NVGRE, in each packet. In the 282 remainder of this document we use VNI as the representation for NVO 283 instance with the understanding that VSID can equally be used if the 284 encapsulation is NVGRE unless it is stated otherwise. 286 Note that a Provider Edge (PE) is equivalent to a NVE/VTEP. 288 VXLAN encapsulation is based on UDP, with an 8-byte header following 289 the UDP header. VXLAN provides a 24-bit VNI, which typically provides 290 a one-to-one mapping to the tenant VLAN ID, as described in 291 [RFC7348]. In this scenario, the ingress VTEP does not include an 292 inner VLAN tag on the encapsulated frame, and the egress VTEP 293 discards the frames with an inner VLAN tag. This mode of operation in 294 [RFC7348] maps to VLAN Based Service in [RFC7432], where a tenant 295 VLAN ID gets mapped to an EVPN instance (EVI). 297 VXLAN also provides an option of including an inner VLAN tag in the 298 encapsulated frame, if explicitly configured at the VTEP. This mode 299 of operation can map to VLAN Bundle Service in [RFC7432] because all 300 the tenant's tagged frames map to a single bridge table / MAC-VRF, 301 and the inner VLAN tag is not used for lookup by the disposition PE 302 when performing VXLAN decapsulation as described in section 6 of 303 [RFC7348]. 305 [RFC7637] encapsulation is based on GRE encapsulation and it mandates 306 the inclusion of the optional GRE Key field which carries the VSID. 307 There is a one-to-one mapping between the VSID and the tenant VLAN 308 ID, as described in [RFC7637] and the inclusion of an inner VLAN tag 309 is prohibited. This mode of operation in [RFC7637] maps to VLAN Based 310 Service in [RFC7432]. 312 As described in the next section there is no change to the encoding 313 of EVPN routes to support VXLAN or NVGRE encapsulation except for the 314 use of the BGP Encapsulation extended community to indicate the 315 encapsulation type (e.g., VxLAN or NVGRE). However, there is 316 potential impact to the EVPN procedures depending on where the NVE is 317 located (i.e., in hypervisor or TOR) and whether multi-homing 318 capabilities are required. 320 5.1.1 Virtual Identifiers Scope 322 Although VNIs are defined as 24-bit globally unique values, there are 323 scenarios in which it is desirable to use a locally significant value 324 for VNI, especially in the context of data center interconnect: 326 5.1.1.1 Data Center Interconnect with Gateway 328 In the case where NVEs in different data centers need to be 329 interconnected, and the NVEs need to use VNIs as a globally unique 330 identifiers within a data center, then a Gateway needs to be employed 331 at the edge of the data center network. This is because the Gateway 332 will provide the functionality of translating the VNI when crossing 333 network boundaries, which may align with operator span of control 334 boundaries. As an example, consider the network of Figure 1 below. 335 Assume there are three network operators: one for each of the DC1, 336 DC2 and WAN networks. The Gateways at the edge of the data centers 337 are responsible for translating the VNIs between the values used in 338 each of the data center networks and the values used in the WAN. 340 +--------------+ 341 | | 342 +---------+ | WAN | +---------+ 343 +----+ | +---+ +----+ +----+ +---+ | +----+ 344 |NVE1|--| | | |WAN | |WAN | | | |--|NVE3| 345 +----+ |IP |GW |--|Edge| |Edge|--|GW | IP | +----+ 346 +----+ |Fabric +---+ +----+ +----+ +---+ Fabric | +----+ 347 |NVE2|--| | | | | |--|NVE4| 348 +----+ +---------+ +--------------+ +---------+ +----+ 350 |<------ DC 1 ------> <------ DC2 ------>| 352 Figure 1: Data Center Interconnect with Gateway 354 5.1.1.2 Data Center Interconnect without Gateway 356 In the case where NVEs in different data centers need to be 357 interconnected, and the NVEs need to use locally assigned VNIs (e.g., 358 similar to MPLS labels), then there may be no need to employ Gateways 359 at the edge of the data center network. More specifically, the VNI 360 value that is used by the transmitting NVE is allocated by the NVE 361 that is receiving the traffic (in other words, this is similar to 362 "downstream assigned" MPLS label). This allows the VNI space to be 363 decoupled between different data center networks without the need for 364 a dedicated Gateway at the edge of the data centers. This topics is 365 covered in section 10.2. 367 +--------------+ 368 | | 369 +---------+ | WAN | +---------+ 370 +----+ | | +----+ +----+ | | +----+ 371 |NVE1|--| | |ASBR| |ASBR| | |--|NVE3| 372 +----+ |IP Fabric|---| | | |--|IP Fabric| +----+ 373 +----+ | | +----+ +----+ | | +----+ 374 |NVE2|--| | | | | |--|NVE4| 375 +----+ +---------+ +--------------+ +---------+ +----+ 377 |<------ DC 1 -----> <---- DC2 ------>| 379 Figure 2: Data Center Interconnect with ASBR 381 5.1.2 Virtual Identifiers to EVI Mapping 383 When the EVPN control plane is used in conjunction with VXLAN (or 384 NVGRE encapsulation), two options for mapping the VXLAN VNI (or NVGRE 385 VSID) to an EVI are possible: 387 1. Option 1: Single Broadcast Domain per EVI 389 In this option, a single Ethernet broadcast domain (e.g., subnet) 390 represented by a VNI is mapped to a unique EVI. This corresponds to 391 the VLAN Based service in [RFC7432], where a tenant-facing interface, 392 logical interface (e.g., represented by a VLAN ID) or physical, gets 393 mapped to an EVPN instance (EVI). As such, a BGP RD and RT are needed 394 per VNI on every NVE. The advantage of this model is that it allows 395 the BGP RT constraint mechanisms to be used in order to limit the 396 propagation and import of routes to only the NVEs that are interested 397 in a given VNI. The disadvantage of this model may be the 398 provisioning overhead if RD and RT are not derived automatically from 399 VNI. 401 In this option, the MAC-VRF table is identified by the RT in the 402 control plane and by the VNI in the data-plane. In this option, the 403 specific MAC-VRF table corresponds to only a single bridge table. 405 2. Option 2: Multiple Broadcast Domains per EVI 407 In this option, multiple subnets each represented by a unique VNI are 408 mapped to a single EVI. For example, if a tenant has multiple 409 segments/subnets each represented by a VNI, then all the VNIs for 410 that tenant are mapped to a single EVI - e.g., the EVI in this case 411 represents the tenant and not a subnet . This corresponds to the 412 VLAN-aware bundle service in [RFC7432]. The advantage of this model 413 is that it doesn't require the provisioning of RD/RT per VNI. 414 However, this is a moot point when compared to option 1 where auto- 415 derivation is used. The disadvantage of this model is that routes 416 would be imported by NVEs that may not be interested in a given VNI. 418 In this option the MAC-VRF table is identified by the RT in the 419 control plane and a specific bridge table for that MAC-VRF is 420 identified by the in the control plane. In this 421 option, the VNI in the data-plane is sufficient to identify a 422 specific bridge table. 424 5.1.2.1 Auto Derivation of RT 426 When the option of a single VNI per EVI is used, in order to simplify 427 configuration, the RT used for EVPN can be auto-derived. RD can be 428 auto generated as described in [RFC7432] and RT can be auto-derived 429 as described next. 431 Since a gateway PE as depicted in figure-1 participates in both the 432 DCN and WAN BGP sessions, it is important that when RT values are 433 auto-derived from VNIs, there is no conflict in RT spaces between DCN 434 and WAN networks assuming that both are operating within the same AS. 435 Also, there can be scenarios where both VXLAN and NVGRE 436 encapsulations may be needed within the same DCN and their 437 corresponding VNIs are administered independently which means VNI 438 spaces can overlap. In order to avoid conflict in RT spaces arises, 439 the 6-byte RT values with 2-octet AS number for DCNs can be auto- 440 derived as follow: 442 0 1 2 3 443 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 444 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 445 | Global Administrator | Local Administrator | 446 +-----------------------------------------------+---------------+ 447 | Local Administrator (Cont.) | 448 +-------------------------------+ 450 0 1 2 3 451 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 452 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 453 | Global Administrator |A| TYPE| D-ID | Service ID | 454 +-----------------------------------------------+---------------+ 455 | Service ID (Cont.) | 456 +-------------------------------+ 458 The 6-octet RT field consists of two sub-field: 460 - Global Administrator sub-field: 2 octets. This sub-field contains 461 an Autonomous System number assigned by IANA. 463 - Local Administrator sub-field: 4 octets 465 * A: A single-bit field indicating if this RT is auto-derived 467 0: auto-derived 468 1: manually-derived 470 * Type: A 3-bit field that identifies the space in which 471 the other 3 bytes are defined. The following spaces are 472 defined: 474 0 : VID (802.1Q VLAN ID) 475 1 : VXLAN 476 2 : NVGRE 477 3 : I-SID 478 4 : EVI 479 5 : dual-VID (QinQ VLAN ID) 481 * D-ID: A 4-bit field that identifies domain-id. The default 482 value of domain-id is zero indicating that only a single 483 numbering space exist for a given technology. However, if 484 there are more than one number space exist for a given 485 technology (e.g., overlapping VXLAN spaces), then each of 486 the number spaces need to be identify by their 487 corresponding domain-id starting from 1. 489 * Service ID: This 3-octet field is set to VNI, VSID, I-SID, 490 or VID. 492 It should be noted that RT auto-derivation is applicable for 2-octet 493 AS numbers. For 4-octet AS numbers, RT needs to be manually 494 configured since 3-octet VNI fields cannot be fit within 2-octet 495 local administrator field. 497 5.1.3 Constructing EVPN BGP Routes 499 In EVPN, an MPLS label for instance identifying forwarding table is 500 distributed by the egress PE via the EVPN control plane and is placed 501 in the MPLS header of a given packet by the ingress PE. This label is 502 used upon receipt of that packet by the egress PE for disposition of 503 that packet. This is very similar to the use of the VNI by the egress 504 NVE, with the difference being that an MPLS label has local 505 significance while a VNI typically has global significance. 506 Accordingly, and specifically to support the option of locally- 507 assigned VNIs, the MPLS Label1 field in the MAC/IP Advertisement 508 route, the MPLS label field in the Ethernet AD per EVI route, and the 509 MPLS label field in the PMSI Tunnel Attribute of the Inclusive 510 Multicast Ethernet Tag (IMET) route are used to carry the VNI. For 511 the balance of this memo, the above MPLS label fields will be 512 referred to as the VNI field. The VNI field is used for both local 513 and global VNIs, and for either case the entire 24-bit field is used 514 to encode the VNI value. 516 For the VLAN-based service (a single VNI per MAC-VRF), the Ethernet 517 Tag field in the MAC/IP Advertisement, Ethernet AD per EVI, and IMET 518 route MUST be set to zero just as in the VLAN Based service in 519 [RFC7432]. 521 For the VLAN-aware bundle service (multiple VNIs per MAC-VRF with 522 each VNI associated with its own bridge table), the Ethernet Tag 523 field in the MAC Advertisement, Ethernet AD per EVI, and IMET route 524 MUST identify a bridge table within a MAC-VRF and the set of Ethernet 525 Tags for that EVI needs to be configured consistently on all PEs 526 within that EVI. For locally-assigned VNIs, the value advertised in 527 the Ethernet Tag field MUST be set to a VID just as in the VLAN-aware 528 bundle service in [RFC7432]. Such setting must be done consistently 529 on all PE devices participating in that EVI within a given domain. 530 For global VNIs, the value advertised in the Ethernet Tag field 531 SHOULD be set to a VNI as long as it matches the existing semantics 532 of the Ethernet Tag, i.e., it identifies a bridge table within a MAC- 533 VRF and the set of VNIs are configured consistently on each PE in 534 that EVI. 536 In order to indicate which type of data plane encapsulation (i.e., 537 VXLAN, NVGRE, MPLS, or MPLS in GRE) is to be used, the BGP 538 Encapsulation extended community defined in [TUNNEL-ENCAP] and 539 [RFC5512] is included with all EVPN routes (i.e. MAC Advertisement, 540 Ethernet AD per EVI, Ethernet AD per ESI, Inclusive Multicast 541 Ethernet Tag, and Ethernet Segment) advertised by an egress PE. Five 542 new values have been assigned by IANA to extend the list of 543 encapsulation types defined in [TUNNEL-ENCAP] and they are listed in 544 section 13. 546 The MPLS encapsulation tunnel type, listed in section 13, is needed 547 in order to distinguish between an advertising node that only 548 supports non-MPLS encapsulations and one that supports MPLS and non- 549 MPLS encapsulations. An advertising node that only supports MPLS 550 encapsulation does not need to advertise any encapsulation tunnel 551 types; i.e., if the BGP Encapsulation extended community is not 552 present, then either MPLS encapsulation or a statically configured 553 encapsulation is assumed. 555 The Next Hop field of the MP_REACH_NLRI attribute of the route MUST 556 be set to the IPv4 or IPv6 address of the NVE. The remaining fields 557 in each route are set as per [RFC7432]. 559 Note that the procedure defined here to use the MPLS Label field to 560 carry the VNI in the presence of a Tunnel Encapsulation Extended 561 Community specifying the use of a VNI, is aligned with the procedures 562 described in section 8.2.2.2 of [TUNNEL-ENCAP] ("When a Valid VNI has 563 not been Signaled"). 565 5.2 MPLS over GRE 567 The EVPN data-plane is modeled as an EVPN MPLS client layer sitting 568 over an MPLS PSN-tunnel server layer. Some of the EVPN functions 569 (split-horizon, aliasing, and backup-path) are tied to the MPLS 570 client layer. If MPLS over GRE encapsulation is used, then the EVPN 571 MPLS client layer can be carried over an IP PSN tunnel transparently. 572 Therefore, there is no impact to the EVPN procedures and associated 573 data-plane operation. 575 The existing standards for MPLS over GRE encapsulation as defined by 576 [RFC4023] can be used for this purpose; however, when it is used in 577 conjunction with EVPN the GRE key field SHOULD be present, and SHOULD 578 be used to provide a 32-bit entropy value. The Checksum and Sequence 579 Number fields MUST NOT be included and the corresponding C and S bits 580 in the GRE Packet Header MUST be set to zero. A PE capable of 581 supporting this encapsulation, should advertise its EVPN routes along 582 with the Tunnel Encapsulation extended community indicating MPLS over 583 GRE encapsulation, as described in previous section. 585 6 EVPN with Multiple Data Plane Encapsulations 587 The use of the BGP Encapsulation extended community per [TUNNEL- 588 ENCAP] and [RFC5512] allows each NVE in a given EVI to know each of 589 the encapsulations supported by each of the other NVEs in that EVI. 590 i.e., each of the NVEs in a given EVI may support multiple data plane 591 encapsulations. An ingress NVE can send a frame to an egress NVE 592 only if the set of encapsulations advertised by the egress NVE forms 593 a non-empty intersection with the set of encapsulations supported by 594 the ingress NVE, and it is at the discretion of the ingress NVE which 595 encapsulation to choose from this intersection. (As noted in 596 section 5.1.3, if the BGP Encapsulation extended community is not 597 present, then the default MPLS encapsulation or a locally configured 598 encapsulation is assumed.) 600 When a PE advertises multiple supported encapsulations, it MUST 601 advertise encapsulations that use the same EVPN procedures including 602 procedures associated with split-horizon filtering described in 603 section 8.3.1. For example, VxLAN and NvGRE (or MPLS and MPLS over 604 GRE) encapsulations use the same EVPN procedures and thus a PE can 605 advertise both of them and can support either of them or both of them 606 simultaneously. However, a PE MUST NOT advertise VxLAN and MPLS 607 encapsulations together because a) MPLS field of EVPN routes is set 608 to either a MPLS label for a VNI but not both and b) some of EVPN 609 procedures (such as split-horizon filtering) are different for 610 VxLAN/NvGRE and MPLS encapsulations. 612 An ingress node that uses shared multicast trees for sending 613 broadcast or multicast frames MAY maintain distinct trees for each 614 different encapsulation type. 616 It is the responsibility of the operator of a given EVI to ensure 617 that all of the NVEs in that EVI support at least one common 618 encapsulation. If this condition is violated, it could result in 619 service disruption or failure. The use of the BGP Encapsulation 620 extended community provides a method to detect when this condition is 621 violated but the actions to be taken are at the discretion of the 622 operator and are outside the scope of this document. 624 7 Single-Homing NVEs - NVE Residing in Hypervisor 626 When a NVE and its hosts/VMs are co-located in the same physical 627 device, e.g., when they reside in a server, the links between them 628 are virtual and they typically share fate; i.e., the subject 629 hosts/VMs are typically not multi-homed or if they are multi-homed, 630 the multi-homing is a purely local matter to the server hosting the 631 VM and the NVEs, and need not be "visible" to any other NVEs residing 632 on other servers, and thus does not require any specific protocol 633 mechanisms. The most common case of this is when the NVE resides on 634 the hypervisor. 636 In the sub-sections that follow, we will discuss the impact on EVPN 637 procedures for the case when the NVE resides on the hypervisor and 638 the VXLAN (or NVGRE) encapsulation is used. 640 7.1 Impact on EVPN BGP Routes & Attributes for VXLAN/NVGRE Encapsulation 642 In scenarios where different groups of data centers are under 643 different administrative domains, and these data centers are 644 connected via one or more backbone core providers as described in 645 [RFC7365], the RD must be a unique value per EVI or per NVE as 646 described in [RFC7432]. In other words, whenever there is more than 647 one administrative domain for global VNI, then a unique RD must be 648 used, or whenever the VNI value has local significance, then a unique 649 RD must be used. Therefore, it is recommended to use a unique RD as 650 described in [RFC7432] at all time. 652 When the NVEs reside on the hypervisor, the EVPN BGP routes and 653 attributes associated with multi-homing are no longer required. This 654 reduces the required routes and attributes to the following subset of 655 four out of the total of eight listed in section 7 of [RFC7432]: 657 - MAC/IP Advertisement Route 658 - Inclusive Multicast Ethernet Tag Route 659 - MAC Mobility Extended Community 660 - Default Gateway Extended Community 662 However, as noted in section 8.6 of [RFC7432] in order to enable a 663 single-homing ingress NVE to take advantage of fast convergence, 664 aliasing, and backup-path when interacting with multi-homed egress 665 NVEs attached to a given Ethernet segment, the single-homing ingress 666 NVE should be able to receive and process Ethernet AD per ES and 667 Ethernet AD per EVI routes. 669 7.2 Impact on EVPN Procedures for VXLAN/NVGRE Encapsulation 671 When the NVEs reside on the hypervisors, the EVPN procedures 672 associated with multi-homing are no longer required. This limits the 673 procedures on the NVE to the following subset of the EVPN procedures: 675 1. Local learning of MAC addresses received from the VMs per section 676 10.1 of [RFC7432]. 678 2. Advertising locally learned MAC addresses in BGP using the MAC/IP 679 Advertisement routes. 681 3. Performing remote learning using BGP per Section 10.2 of 682 [RFC7432]. 684 4. Discovering other NVEs and constructing the multicast tunnels 685 using the Inclusive Multicast Ethernet Tag routes. 687 5. Handling MAC address mobility events per the procedures of Section 688 16 in [RFC7432]. 690 However, as noted in section 8.6 of [RFC7432] in order to enable a 691 single-homing ingress NVE to take advantage of fast convergence, 692 aliasing, and back-up path when interacting with multi-homed egress 693 NVEs attached to a given Ethernet segment, a single-homing ingress 694 NVE should implement the ingress node processing of Ethernet AD per 695 ES and Ethernet AD per EVI routes as defined in sections 8.2 Fast 696 Convergence and 8.4 Aliasing and Backup-Path of [RFC7432]. 698 8 Multi-Homing NVEs - NVE Residing in ToR Switch 700 In this section, we discuss the scenario where the NVEs reside in the 701 Top of Rack (ToR) switches AND the servers (where VMs are residing) 702 are multi-homed to these ToR switches. The multi-homing NVE operate 703 in All-Active or Single-Active redundancy mode. If the servers are 704 single-homed to the ToR switches, then the scenario becomes similar 705 to that where the NVE resides on the hypervisor, as discussed in 706 Section 7, as far as the required EVPN functionality are concerned. 708 [RFC7432] defines a set of BGP routes, attributes and procedures to 709 support multi-homing. We first describe these functions and 710 procedures, then discuss which of these are impacted by the VxLAN 711 (or NVGRE) encapsulation and what modifications are required. As it 712 will be seen later in this section, the only EVPN procedure that is 713 impacted by non-MPLS overlay encapsulation (e.g., VxLAN or NVGRE) 714 where it provides space for one ID rather than stack of labels, is 715 that of split-horizon filtering for multi-homed Ethernet Segments 716 described in section 8.3.1. 718 8.1 EVPN Multi-Homing Features 720 In this section, we will recap the multi-homing features of EVPN to 721 highlight the encapsulation dependencies. The section only describes 722 the features and functions at a high-level. For more details, the 723 reader is to refer to [RFC7432]. 725 8.1.1 Multi-homed Ethernet Segment Auto-Discovery 727 EVPN NVEs (or PEs) connected to the same Ethernet Segment (e.g. the 728 same server via LAG) can automatically discover each other with 729 minimal to no configuration through the exchange of BGP routes. 731 8.1.2 Fast Convergence and Mass Withdraw 733 EVPN defines a mechanism to efficiently and quickly signal, to remote 734 NVEs, the need to update their forwarding tables upon the occurrence 735 of a failure in connectivity to an Ethernet segment (e.g., a link or 736 a port failure). This is done by having each NVE advertise an 737 Ethernet A-D Route per Ethernet segment for each locally attached 738 segment. Upon a failure in connectivity to the attached segment, the 739 NVE withdraws the corresponding Ethernet A-D route. This triggers all 740 NVEs that receive the withdrawal to update their next-hop adjacencies 741 for all MAC addresses associated with the Ethernet segment in 742 question. If no other NVE had advertised an Ethernet A-D route for 743 the same segment, then the NVE that received the withdrawal simply 744 invalidates the MAC entries for that segment. Otherwise, the NVE 745 updates the next-hop adjacency list accordingly. 747 8.1.3 Split-Horizon 749 If a server is multi-homed to two or more NVEs (represented by an 750 Ethernet segment ES1) and operating in an all-active redundancy mode, 751 sends a BUM packet (ie, Broadcast, Unknown unicast, or Multicast) to 752 one of these NVEs, then it is important to ensure the packet is not 753 looped back to the server via another NVE connected to this server. 754 The filtering mechanism on the NVE to prevent such loop and packet 755 duplication is called "split horizon filtering'. 757 8.1.4 Aliasing and Backup-Path 759 In the case where a station is multi-homed to multiple NVEs, it is 760 possible that only a single NVE learns a set of the MAC addresses 761 associated with traffic transmitted by the station. This leads to a 762 situation where remote NVEs receive MAC advertisement routes, for 763 these addresses, from a single NVE even though multiple NVEs are 764 connected to the multi-homed station. As a result, the remote NVEs 765 are not able to effectively load-balance traffic among the NVEs 766 connected to the multi-homed Ethernet segment. This could be the 767 case, for e.g. when the NVEs perform data-path learning on the 768 access, and the load-balancing function on the station hashes traffic 769 from a given source MAC address to a single NVE. Another scenario 770 where this occurs is when the NVEs rely on control plane learning on 771 the access (e.g. using ARP), since ARP traffic will be hashed to a 772 single link in the LAG. 774 To alleviate this issue, EVPN introduces the concept of Aliasing. 775 This refers to the ability of an NVE to signal that it has 776 reachability to a given locally attached Ethernet segment, even when 777 it has learnt no MAC addresses from that segment. The Ethernet A-D 778 route per EVI is used to that end. Remote NVEs which receive MAC 779 advertisement routes with non-zero ESI should consider the MAC 780 address as reachable via all NVEs that advertise reachability to the 781 relevant Segment using Ethernet A-D routes with the same ESI and with 782 the Single-Active flag reset. 784 Backup-Path is a closely related function, albeit it applies to the 785 case where the redundancy mode is Single-Active. In this case, the 786 NVE signals that it has reachability to a given locally attached 787 Ethernet Segment using the Ethernet A-D route as well. Remote NVEs 788 which receive the MAC advertisement routes, with non-zero ESI, should 789 consider the MAC address as reachable via the advertising NVE. 790 Furthermore, the remote NVEs should install a Backup-Path, for said 791 MAC, to the NVE which had advertised reachability to the relevant 792 Segment using an Ethernet A-D route with the same ESI and with the 793 Single-Active flag set. 795 8.1.5 DF Election 797 If a host is multi-homed to two or more NVEs on an Ethernet segment 798 operating in all-active redundancy mode, then for a given EVI only 799 one of these NVEs, termed the Designated Forwarder (DF) is 800 responsible for sending it broadcast, multicast, and, if configured 801 for that EVI, unknown unicast frames. 803 This is required in order to prevent duplicate delivery of multi- 804 destination frames to a multi-homed host or VM, in case of all-active 805 redundancy. 807 In NVEs where .1Q tagged frames are received from hosts, the DF 808 election should be performed based on host VLAN IDs (VIDs) per 809 section 8.5 of [RFC7432]. Furthermore, multi-homing PEs of a given 810 Ethernet Segment MAY perform DF election using configured IDs such as 811 VNI, EVI, normalized VIDs, and etc. as along the IDs are configured 812 consistently across the multi-homing PEs. 814 In GWs where VxLAN encapsulated frames are received, the DF election 815 is performed on VNIs. Again, it is assumed that for a given Ethernet 816 Segment, VNIs are unique and consistent (e.g., no duplicate VNIs 817 exist). 819 8.2 Impact on EVPN BGP Routes & Attributes 821 Since multi-homing is supported in this scenario, then the entire set 822 of BGP routes and attributes defined in [RFC7432] are used. The 823 setting of the Ethernet Tag field in the MAC Advertisement, Ethernet 824 AD per EVI, and Inclusive Multicast routes follows that of section 825 5.1.3. Furthermore, the setting of the VNI field in the MAC 826 Advertisement and Ethernet AD per EVI routes follows that of section 827 5.1.3. 829 8.3 Impact on EVPN Procedures 831 Two cases need to be examined here, depending on whether the NVEs are 832 operating in Single-Active or in All-Active redundancy mode. 834 First, lets consider the case of Single-Active redundancy mode, where 835 the hosts are multi-homed to a set of NVEs, however, only a single 836 NVE is active at a given point of time for a given VNI. In this case, 837 the aliasing is not required and the split-horizon filtering may not 838 be required, but other functions such as multi-homed Ethernet segment 839 auto-discovery, fast convergence and mass withdraw, backup path, and 840 DF election are required. 842 Second, let's consider the case of All-Active redundancy mode. In 843 this case, out of all the EVPN multi-homing features listed in 844 section 8.1, the use of the VXLAN or NVGRE encapsulation impacts the 845 split-horizon and aliasing features, since those two rely on the MPLS 846 client layer. Given that this MPLS client layer is absent with these 847 types of encapsulations, alternative procedures and mechanisms are 848 needed to provide the required functions. Those are discussed in 849 detail next. 851 8.3.1 Split Horizon 853 In EVPN, an MPLS label is used for split-horizon filtering to support 854 All-Active multi-homing where an ingress NVE adds a label 855 corresponding to the site of origin (aka ESI Label) when 856 encapsulating the packet. The egress NVE checks the ESI label when 857 attempting to forward a multi-destination frame out an interface, and 858 if the label corresponds to the same site identifier (ESI) associated 859 with that interface, the packet gets dropped. This prevents the 860 occurrence of forwarding loops. 862 Since VXLAN and NVGRE encapsulations do not include the ESI label, 863 other means of performing the split-horizon filtering function must 864 be devised for these encapsulations. The following approach is 865 recommended for split-horizon filtering when VXLAN (or NVGRE) 866 encapsulation is used. 868 Every NVE track the IP address(es) associated with the other NVE(s) 869 with which it has shared multi-homed Ethernet Segments. When the NVE 870 receives a multi-destination frame from the overlay network, it 871 examines the source IP address in the tunnel header (which 872 corresponds to the ingress NVE) and filters out the frame on all 873 local interfaces connected to Ethernet Segments that are shared with 874 the ingress NVE. With this approach, it is required that the ingress 875 NVE performs replication locally to all directly attached Ethernet 876 Segments (regardless of the DF Election state) for all flooded 877 traffic ingress from the access interfaces (i.e. from the hosts). 878 This approach is referred to as "Local Bias", and has the advantage 879 that only a single IP address needs to be used per NVE for split- 880 horizon filtering, as opposed to requiring an IP address per Ethernet 881 Segment per NVE. 883 In order to allow proper operation of split-horizon filtering among 884 the same group of multi-homing PE devices, a mix of PE devices with 885 MPLS over GRE encapsulations running [RFC7432] procedures for split- 886 horizon filtering on the one hand and VXLAN/NVGRE encapsulations 887 running local-bias procedures on the other on a given Ethernet 888 Segment MUST NOT be configured. 890 8.3.2 Aliasing and Backup-Path 892 The Aliasing and the Backup-Path procedures for VXLAN/NVGRE 893 encapsulation are very similar to the ones for MPLS. In case of MPLS, 894 Ethernet A-D route per EVI is used for Aliasing when the 895 corresponding Ethernet Segment operates in All-Active multi-homing, 896 and the same route is used for Backup-Path when the corresponding 897 Ethernet Segment operates in Single-Active multi-homing. In case of 898 VxLAN/NVGRE, the same route is used for the Aliasing and the Backup- 899 Path with the difference that the Ethernet Tag and VNI fields in 900 Ethernet A-D per EVI route are set as described in section 5.1.3. 902 8.3.3 Unknown Unicast Traffic Designation 904 In EVPN, when an ingress PE uses ingress replication to flood unknown 905 unicast traffic to egress PEs, the ingress PE uses a different EVPN 906 MPLS label (from the one used for known unicast traffic) to identify 907 such BUM traffic. The egress PEs use this label to identify such BUM 908 traffic and thus apply DF filtering for All-Active multi-homed sites. 909 In absence of unknown unicast traffic designation and in presence of 910 enabling unknown unicast flooding, there can be transient duplicate 911 traffic to All-Active multi-homed sites under the following 912 condition: the host MAC address is learned by the egress PE(s) and 913 advertised to the ingress PE; however, the MAC advertisement has not 914 been received or processed by the ingress PE, resulting in the host 915 MAC address to be unknown on the ingress PE but be known on the 916 egress PE(s). Therefore, when a packet destined to that host MAC 917 address arrives on the ingress PE, it floods it via ingress 918 replication to all the egress PE(s) and since they are known to the 919 egress PE(s), multiple copies is sent to the All-Active multi-homed 920 site. It should be noted that such transient packet duplication only 921 happens when a) the destination host is multi-homed via All-Active 922 redundancy mode, b) flooding of unknown unicast is enabled in the 923 network, c) ingress replication is used, and d) traffic for the 924 destination host is arrived on the ingress PE before it learns the 925 host MAC address via BGP EVPN advertisement. In order to prevent such 926 occurrence of packet duplication (however low probability that may 927 be), the ingress PE MAY set the BUM Traffic Bit (B bit) [VXLAN-GPE] 928 to indicate BUM traffic. 930 9 Support for Multicast 932 The E-VPN Inclusive Multicast Ethernet Tag (IMET) route is used to 933 discover the multicast tunnels among the endpoints associated with a 934 given EVI (e.g., given VNI) for VLAN-based service and a given 935 for VLAN-aware bundle service. All fields of this route is 936 set as described in section 5.1.3. The Originating router's IP 937 address field is set to the NVE's IP address. This route is tagged 938 with the PMSI Tunnel attribute, which is used to encode the type of 939 multicast tunnel to be used as well as the multicast tunnel 940 identifier. The tunnel encapsulation is encoded by adding the BGP 941 Encapsulation extended community as per section 5.1.1. For example, 942 the PMSI Tunnel attribute may indicate the multicast tunnel is of 943 type PIM-SM; whereas, the BGP Encapsulation extended community may 944 indicate the encapsulation for that tunnel is of type VxLAN. The 945 following tunnel types as defined in [RFC6514] can be used in the 946 PMSI tunnel attribute for VXLAN/NVGRE: 948 + 3 - PIM-SSM Tree 949 + 4 - PIM-SM Tree 950 + 5 - BIDIR-PIM Tree 951 + 6 - Ingress Replication 953 Except for Ingress Replication, this multicast tunnel is used by the 954 PE originating the route for sending multicast traffic to other PEs, 955 and is used by PEs that receive this route for receiving the traffic 956 originated by hosts connected to the PE that originated the route. 958 In the scenario where the multicast tunnel is a tree, both the 959 Inclusive as well as the Aggregate Inclusive variants may be used. In 960 the former case, a multicast tree is dedicated to a VNI. Whereas, in 961 the latter, a multicast tree is shared among multiple VNIs. For VNI- 962 based service, the Aggregate Inclusive mode is accomplished by having 963 the NVEs advertise multiple IMET routes with different Route Targets 964 (one per VNI) but with the same tunnel identifier encoded in the PMSI 965 tunnel attribute. For VNI-aware bundle service, the Aggregate 966 Inclusive mode is accomplished by having the NVEs advertise multiple 967 IMET routes with different VNI encoded in the Ethernet Tag field, but 968 with the same tunnel identifier encoded in the PMSI Tunnel attribute. 970 10 Data Center Interconnections - DCI 972 For DCI, the following two main scenarios are considered when 973 connecting data centers running evpn-overlay (as described here) over 974 MPLS/IP core network: 976 - Scenario 1: DCI using GWs 977 - Scenario 2: DCI using ASBRs 979 The following two subsections describe the operations for each of 980 these scenarios. 982 10.1 DCI using GWs 984 This is the typical scenario for interconnecting data centers over 985 WAN. In this scenario, EVPN routes are terminated and processed in 986 each GW and MAC/IP routes are always re-advertised from DC to WAN but 987 from WAN to DC, they are not re-advertised if unknown MAC address 988 (and default IP address) are utilized in NVEs. In this scenario, each 989 GW maintains a MAC-VRF (and/or IP-VRF) for each EVI. The main 990 advantage of this approach is that NVEs do not need to maintain MAC 991 and IP addresses from any remote data centers when default IP route 992 and unknown MAC routes are used - i.e., they only need to maintain 993 routes that are local to their own DC. When default IP route and 994 unknown MAC route are used, any unknown IP and MAC packets from NVEs 995 are forwarded to the GWs where all the VPN MAC and IP routes are 996 maintained. This approach reduces the size of MAC-VRF and IP-VRF 997 significantly at NVEs. Furthermore, it results in a faster 998 convergence time upon a link or NVE failure in a multi-homed network 999 or device redundancy scenario, because the failure related BGP routes 1000 (such as mass withdraw message) do not need to get propagated all the 1001 way to the remote NVEs in the remote DCs. This approach is described 1002 in details in section 3.4 of [DCI-EVPN-OVERLAY]. 1004 10.2 DCI using ASBRs 1006 This approach can be considered as the opposite of the first approach 1007 and it favors simplification at DCI devices over NVEs such that 1008 larger MAC-VRF (and IP-VRF) tables need to be maintained on NVEs; 1009 whereas, DCI devices don't need to maintain any MAC (and IP) 1010 forwarding tables. Furthermore, DCI devices do not need to terminate 1011 and process routes related to multi-homing but rather to relay these 1012 messages for the establishment of an end-to-end LSP path. In other 1013 words, DCI devices in this approach operate similar to ASBRs for 1014 inter-AS option B - section 10 of [RFC4364]. This requires locally 1015 assigned VNIs to be used just like downstream assigned MPLS VPN label 1016 where for all practical purposes the VNIs function like 24-bit VPN 1017 labels. This approach is equally applicable to data centers (or 1018 Carrier Ethernet networks) with MPLS encapsulation. 1020 In inter-AS option B, when ASBR receives an EVPN route from its DC 1021 over iBGP and re-advertises it to other ASBRs, it re-advertises the 1022 EVPN route by re-writing the BGP next-hops to itself, thus losing the 1023 identity of the PE that originated the advertisement. This re-write 1024 of BGP next-hop impacts the EVPN Mass Withdraw route (Ethernet A-D 1025 per ES) and its procedure adversely. However, it does not impact EVPN 1026 Aliasing mechanism/procedure because when the Aliasing routes (Ether 1027 A-D per EVI) are advertised, the receiving PE first resolves a MAC 1028 address for a given EVI into its corresponding and 1029 subsequently, it resolves the into multiple paths (and their 1030 associated next hops) via which the is reachable. Since 1031 Aliasing and MAC routes are both advertised per EVI basis and they 1032 use the same RD and RT (per EVI), the receiving PE can associate them 1033 together on a per BGP path basis (e.g., per originating PE) and thus 1034 perform recursive route resolution - e.g., a MAC is reachable via an 1035 which in turn, is reachable via a set of BGP paths, thus the 1036 MAC is reachable via the set of BGP paths. Since on a per EVI basis, 1037 the association of MAC routes and the corresponding Aliasing route is 1038 fixed and determined by the same RD and RT, there is no ambiguity 1039 when the BGP next hop for these routes is re-written as these routes 1040 pass through ASBRs - i.e., the receiving PE may receive multiple 1041 Aliasing routes for the same EVI from a single next hop (a single 1042 ASBR), and it can still create multiple paths toward that . 1044 However, when the BGP next hop address corresponding to the 1045 originating PE is re-written, the association between the Mass 1046 Withdraw route (Ether A-D per ES) and its corresponding MAC routes 1047 cannot be made based on their RDs and RTs because the RD for Mass 1048 Withdraw route is different than the one for the MAC routes. 1049 Therefore, the functionality needed at the ASBRs and the receiving 1050 PEs depends on whether the Mass Withdraw route is originated and 1051 whether there is a need to handle route resolution ambiguity for this 1052 route. The following two subsections describe the functionality 1053 needed by the ASBRs and the receiving PEs depending on whether the 1054 NVEs reside in a Hypervisors or in TORs. 1056 10.2.1 ASBR Functionality with Single-Homing NVEs 1058 When NVEs reside in hypervisors as described in section 7.1, there is 1059 no multi-homing and thus there is no need for the originating NVE to 1060 send Ethernet A-D per ES or Ethernet A-D per EVI routes. However, as 1061 noted in section 7, in order to enable a single-homing ingress NVE to 1062 take advantage of fast convergence, aliasing, and backup-path when 1063 interacting with multi-homing egress NVEs attached to a given 1064 Ethernet segment, the single-homing NVE should be able to receive and 1065 process Ethernet AD per ES and Ethernet AD per EVI routes. The 1066 handling of these routes are described in the next section. 1068 10.2.2 ASBR Functionality with Multi-Homing NVEs 1070 When NVEs reside in TORs and operate in multi-homing redundancy mode, 1071 then as described in section 8, there is a need for the originating 1072 multi-homing NVE to send Ethernet A-D per ES route(s) (used for mass 1073 withdraw) and Ethernet A-D per EVI routes (used for aliasing). As 1074 described above, the re-write of BGP next-hop by ASBRs creates 1075 ambiguities when Ethernet A-D per ES routes are received by the 1076 remote NVE in a different ASBR because the receiving NVE cannot 1077 associated that route with the MAC/IP routes of that Ethernet Segment 1078 advertised by the same originating NVE. This ambiguity inhibits the 1079 function of mass-withdraw per ES by the receiving NVE in a different 1080 AS. 1082 As an example consider a scenario where CE is multi-homed to PE1 and 1083 PE2 where these PEs are connected via ASBR1 and then ASBR2 to the 1084 remote PE3. Furthermore, consider that PE1 receives M1 from CE1 but 1085 not PE2. Therefore, PE1 advertises Eth A-D per ES1, Eth A-D per EVI1, 1086 and M1; whereas, PE2 only advertises Eth A-D per ES1 and Eth A-D per 1087 EVI1. ASBR1 receives all these five advertisements and passes them to 1088 ASBR2 (with itself as the BGP next hop). ASBR2, in turn, passes them 1089 to the remote PE3 with itself as the BGP next hop. PE3 receives these 1090 five routes where all of them have the same BGP next-hop (i.e., 1091 ASBR2). Furthermore, the two Ether A-D per ES routes received by PE3 1092 have the same info - i.e., same ESI and the same BGP next hop. 1093 Although both of these routes are maintained by the BGP process in 1094 PE3 (because they have different RDs and thus treated as different 1095 BGP routes), information from only one of them is used in the L2 1096 routing table (L2 RIB). 1098 PE1 1099 / \ 1100 CE ASBR1---ASBR2---PE3 1101 \ / 1102 PE2 1104 Figure 1: Inter-AS Option B 1106 Now, when the AC between the PE2 and the CE fails and PE2 sends NLRI 1107 withdrawal for Ether A-D per ES route and this withdrawal gets 1108 propagated and received by the PE3, the BGP process in PE3 removes 1109 the corresponding BGP route; however, it doesn't remove the 1110 associated info (namely ESI and BGP next hop) from the L2 routing 1111 table (L2 RIB) because it still has the other Ether A-D per ES route 1112 (originated from PE1) with the same info. That is why the mass- 1113 withdraw mechanism does not work when doing DCI with inter-AS option 1114 B. However, as described previoulsy, the aliasing function works and 1115 so does "mass-withdraw per EVI" (which is associated with withdrawing 1116 the EVPN route associated with Aliasing - i.e., Ether A-D per EVI 1117 route). 1119 In the above example, the PE3 receives two Aliasing routes with the 1120 same BGP next hop (ASBR2) but different RDs. One of the Alias route 1121 has the same RD as the advertised MAC route (M1). PE3 follows the 1122 route resolution procedure specified in [RFC7432] upon receiving the 1123 two Aliasing route - ie, it resolves M1 to and 1124 subsequently it resolves to a BGP path list with two paths 1125 along with the corresponding VNIs/MPLS labels (one associated with 1126 PE1 and the other associated with PE2). It should be noted that even 1127 though both paths are advertised by the same BGP next hop (ASRB2), 1128 the receiving PE3 can handle them properly. Therefore, M1 is 1129 reachable via two paths. This creates two end-to-end LSPs, from PE3 1130 to PE1 and from PE3 to PE2, for M1 such that when PE3 wants to 1131 forward traffic destined to M1, it can load balanced between the two 1132 LSPs. Although route resolution for Aliasing routes with the same BGP 1133 next hop is not explicitly mentioned in [RFC7432], this is the 1134 expected operation and thus it is elaborated here. 1136 When the AC between the PE2 and the CE fails and PE2 sends NLRI 1137 withdrawal for Ether A-D per EVI routes and these withdrawals get 1138 propagated and received by the PE3, the PE3 removes the Aliasing 1139 route and updates the path list - ie, it removes the path 1140 corresponding to the PE2. Therefore, all the corresponding MAC routes 1141 for that that point to that path list will now have the 1142 updated path list with a single path associated with PE1. This action 1143 can be considered as the mass-withdraw at the per-EVI level. The 1144 mass-withdraw at per-EVI level has longer convergence time than the 1145 mass-withdraw at per-ES level; however, it is much faster than the 1146 convergence time when the withdraw is done on a per-MAC basis. 1148 If a PE becomes detached from a given ES, then in addition to 1149 withdrawing its previously advertised Ethernet AD Per ES routes, it 1150 MUST also withdraw its previously advertised Ethernet AD Per EVI 1151 routes for that ES. For a remote PE that is separated from the 1152 withdrawing PE by one or more EVPN inter-AS option B ASBRs, the 1153 withdrawal of the Ethernet AD Per ES routes is not actionable. 1154 However, a remote PE is able to correlate a previously advertised 1155 Ethernet AD Per EVI route with any MAC/IP Advertisement routes also 1156 advertised by the withdrawing PE for that . Hence, when 1157 it receives the withdrawal of an Ethernet AD Per EVI route, it SHOULD 1158 remove the withdrawing PE as a next-hop for all MAC addresses 1159 associated with that . 1161 In the previous example, when the AC between PE2 and the CE fails, 1162 PE2 will withdraw its Ethernet AD Per ES and Per EVI routes. When 1163 PE3 receives the withdrawal of an Ethernet AD Per EVI route, it 1164 removes PE2 as a valid next-hop for all MAC addresses associated with 1165 the corresponding . Therefore, all the MAC next-hops 1166 for that will now have a single next-hop, viz the LSP to 1167 PE1. 1169 In summary, it can be seen that aliasing (and backup path) 1170 functionality should work as is for inter-AS option B without 1171 requiring any addition functionality in ASBRs or PEs. However, the 1172 mass-withdraw functionality falls back from per-ES mode to per-EVI 1173 mode for inter-AS option B - i.e., PEs receiving mass-withdraw route 1174 from the same AS take action on Ether A-D per ES route; whereas, PEs 1175 receiving mass-withdraw route from different AS take action on Ether 1176 A-D per EVI route. 1178 11 Acknowledgement 1180 The authors would like to thank Aldrin Isaac, David Smith, John 1181 Mullooly, Thomas Nadeau, Samir Thoria, and Jorge Rabadan for their 1182 valuable comments and feedback. The authors would also like to thank 1183 Jakob Heitz for his contribution on section 10.2. 1185 12 Security Considerations 1187 This document uses IP-based tunnel technologies to support data 1188 plane transport. Consequently, the security considerations of those 1189 tunnel technologies apply. This document defines support for VXLAN 1190 [RFC7348] and NVGRE [RFC7637] encapsulations. The security 1191 considerations from those RFCs apply to the data plane aspects of 1192 this document. 1194 As with [RFC5512], any modification of the information that is used 1195 to form encapsulation headers, to choose a tunnel type, or to choose 1196 a particular tunnel for a particular payload type may lead to user 1197 data packets getting misrouted, misdelivered, and/or dropped. 1199 More broadly, the security considerations for the transport of IP 1200 reachability information using BGP are discussed in [RFC4271] and 1201 [RFC4272], and are equally applicable for the extensions described 1202 in this document. 1204 13 IANA Considerations 1206 IANA has allocated the following BGP Tunnel Encapsulation Attribute 1207 Tunnel Types: 1209 8 VXLAN Encapsulation 1210 9 NVGRE Encapsulation 1211 10 MPLS Encapsulation 1212 11 MPLS in GRE Encapsulation 1213 12 VXLAN GPE Encapsulation 1215 14 References 1217 14.1 Normative References 1219 [KEYWORDS] Bradner, S., "Key words for use in RFCs to Indicate 1220 Requirement Levels", BCP 14, RFC 2119, March 1997. 1222 [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 1223 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, 1224 May 2017, . 1226 [RFC5512] Mohapatra, P. and E. Rosen, "The BGP Encapsulation 1227 Subsequent Address Family Identifier (SAFI) and the BGP 1228 Tunnel Encapsulation Attribute", RFC 5512, April 2009. 1230 [RFC7432] Sajassi et al., "BGP MPLS Based Ethernet VPN", RFC 7432, 1231 February 2014 1233 14.2 Informative References 1235 [RFC7209] Sajassi et al., "Requirements for Ethernet VPN (EVPN)", RFC 1236 7209, May 2014 1238 [RFC7348] Mahalingam, M., et al, "VXLAN: A Framework for Overlaying 1239 Virtualized Layer 2 Networks over Layer 3 Networks", RFC 7348, August 1240 2014 1242 [RFC4272] S. Murphy, "BGP Security Vulnerabilities Analysis.", 1243 January 2006. 1245 [RFC7637] Garg, P., et al., "NVGRE: Network Virtualization using 1246 Generic Routing Encapsulation", RFC 7637, September, 2015 1248 [RFC7364] Narten et al., "Problem Statement: Overlays for Network 1249 Virtualization", RFC 7364, October 2014. 1251 [RFC7365] Lasserre et al., "Framework for DC Network Virtualization", 1252 RFC 7365, October 2014. 1254 [DCI-EVPN-OVERLAY] Rabadan et al., "Interconnect Solution for EVPN 1255 Overlay networks", draft-ietf-bess-dci-evpn-overlay-04, work in 1256 progress, February 29, 2016. 1258 [RFC4271] Y. Rekhter, Ed., T. Li, Ed., S. Hares, Ed., "A Border 1259 Gateway Protocol 4 (BGP-4)", January 2006. 1261 [TUNNEL-ENCAP] Rosen et al., "The BGP Tunnel Encapsulation 1262 Attribute", draft-ietf-idr-tunnel-encaps-03, work in progress, May 1263 31, 2016. 1265 [VXLAN-GPE] Maino et al., "Generic Protocol Extension for VXLAN", 1266 draft-ietf-nvo3-vxlan-gpe-03, work in progress October 25, 2016. 1268 [RFC4364] Rosen, E., et al, "BGP/MPLS IP Virtual Private Networks 1269 (VPNs)", RFC 4364, February 2006. 1271 [RFC4023] T. Worster et al., "Encapsulating MPLS in IP or Generic 1272 Routing Encapsulation (GRE)", RFC 4023, March 2005 1274 [RFC6514] R. Aggarwal et al., "BGP Encodings and Procedures for 1275 Multicast in MPLS/BGP IP VPNs", RFC 6514, February 2012 1277 Contributors 1279 S. Salam 1280 K. Patel 1281 D. Rao 1282 S. Thoria 1283 D. Cai 1284 Cisco 1286 Y. Rekhter 1287 A. Issac 1288 Wen Lin 1289 Nischal Sheth 1290 Juniper 1292 L. Yong 1293 Huawei 1295 Authors' Addresses 1297 Ali Sajassi 1298 Cisco 1299 Email: sajassi@cisco.com 1301 John Drake 1302 Juniper Networks 1303 Email: jdrake@juniper.net 1305 Nabil Bitar 1306 Nokia 1307 Email : nabil.bitar@nokia.com 1309 R. Shekhar 1310 Juniper 1311 Email: rshekhar@juniper.net 1313 James Uttaro 1314 AT&T 1315 Email: uttaro@att.com 1317 Wim Henderickx 1318 Alcatel-Lucent 1319 e-mail: wim.henderickx@nokia.com