idnits 2.17.1 draft-ietf-rtgwg-net2cloud-problem-statement-02.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 are 6 instances of too long lines in the document, the longest one being 109 characters in excess of 72. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (June 17, 2019) is 1768 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Unused Reference: 'ITU-T-X1036' is defined on line 656, but no explicit reference was found in the text Summary: 1 error (**), 0 flaws (~~), 2 warnings (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 1 Network Working Group L. Dunbar 2 Internet Draft A. Malis 3 Intended status: Informational Futurewei 4 Expires: Dec 2019 C. Jacquenet 5 Orange 6 M. Toy 7 Verizon 8 June 17, 2019 10 Dynamic Networks to Hybrid Cloud DCs Problem Statement 11 draft-ietf-rtgwg-net2cloud-problem-statement-02 13 Abstract 15 This document describes the problems that enterprises face today 16 when connecting their branch offices to dynamic workloads in third 17 party data centers (a.k.a. Cloud DCs). 19 It examines some of the approaches interconnecting cloud DCs with 20 enterprises' on-premises DCs & branch offices. This document also 21 describes some of the network problems that many enterprises face 22 when they have workloads & applications & data split among hybrid 23 data centers, especially for those enterprises with multiple sites 24 that are already interconnected by VPNs (e.g., MPLS L2VPN/L3VPN). 26 Current operational problems are examined to determine whether there 27 is a need to improve existing protocols or whether a new protocol is 28 necessary to solve them. 30 Status of this Memo 32 This Internet-Draft is submitted in full conformance with the 33 provisions of BCP 78 and BCP 79. 35 Internet-Drafts are working documents of the Internet Engineering 36 Task Force (IETF), its areas, and its working groups. Note that 37 other groups may also distribute working documents as Internet- 38 Drafts. 40 Internet-Drafts are draft documents valid for a maximum of six 41 months and may be updated, replaced, or obsoleted by other documents 42 at any time. It is inappropriate to use Internet-Drafts as 43 reference material or to cite them other than as "work in progress." 45 The list of current Internet-Drafts can be accessed at 46 http://www.ietf.org/ietf/1id-abstracts.txt 48 The list of Internet-Draft Shadow Directories can be accessed at 49 http://www.ietf.org/shadow.html 51 This Internet-Draft will expire on December 17, 2019. 53 Copyright Notice 55 Copyright (c) 2019 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 63 respect to this document. Code Components extracted from this 64 document must include Simplified BSD License text as described in 65 Section 4.e of the Trust Legal Provisions and are provided without 66 warranty as described in the Simplified BSD License. 68 Table of Contents 70 1. Introduction...................................................3 71 1.1. On the evolution of Cloud DC connectivity.................3 72 1.2. The role of SD-WAN techniques in Cloud DC connectivity....4 73 2. Definition of terms............................................4 74 3. Current Practices in Interconnecting Enterprise Sites with Cloud 75 DCs...............................................................5 76 3.1. Multiple connection to workloads in a Cloud DC............5 77 3.2. Interconnect to Hybrid Cloud DCs..........................7 78 3.3. Connecting workloads among hybrid Cloud DCs...............8 79 4. Desired Properties for Networks that interconnect Hybrid Clouds9 80 5. Problems with MPLS-based VPNs extending to Hybrid Cloud DCs....9 81 6. Problem with using IPsec tunnels to Cloud DCs.................11 82 6.1. Complexity of multi-point any-to-any interconnection.....11 83 6.2. Poor performance over long distance......................12 84 6.3. Scaling Issues with IPsec Tunnels........................12 86 7. Problems of Using SD-WAN to connect to Cloud DCs..............13 87 7.1. SD-WAN among branch offices vs. interconnect to Cloud DCs13 88 8. End-to-End Security Concerns for Data Flows...................16 89 9. Requirements for Dynamic Cloud Data Center VPNs...............16 90 10. Security Considerations......................................17 91 11. IANA Considerations..........................................17 92 12. References...................................................17 93 12.1. Normative References....................................17 94 12.2. Informative References..................................17 95 13. Acknowledgments..............................................18 97 1. Introduction 99 1.1. On the evolution of Cloud DC connectivity 101 The ever-increasing use of cloud applications for communication 102 services change the way corporate business works and shares 103 information. Such cloud applications use resources hosted in third 104 party DCs that also host services for other customers. 106 With the advent of widely available third party cloud DCs in diverse 107 geographic locations and the advancement of tools for monitoring and 108 predicting application behaviors, it is technically feasible for 109 enterprises to instantiate applications and workloads in locations 110 that are geographically closest to their end-users. Such proximity 111 improves end-to-end latency and overall user experience. Conversely, 112 an enterprise can easily shutdown applications and workloads 113 whenever end-users are in motion (thereby modifying the networking 114 connection of subsequently relocated applications and workloads). In 115 addition, an enterprise may wish to take advantage of more and more 116 business applications offered by third party private cloud DCs. 118 Most of those enterprise branch offices & on-premises data centers 119 are already connected via VPNs, such as MPLS-based L2VPNs and 120 L3VPNs. Then connecting to the cloud-hosted resources may not be 121 straightforward if the provider of the VPN service does not have 122 direct connections to the corresponding cloud DCs. Under those 123 circumstances, the enterprise can upgrade the CPEs deployed in its 124 various premises to utilize SD-WAN techniques to reach cloud 125 resources (without any assistance from the VPN service provider), or 126 wait for their VPN service provider to make new agreements with data 127 center providers to connect to the cloud resources. Either way has 128 additional infrastructure and operational costs. 130 In addition, more enterprises are moving towards hybrid cloud DCs, 131 i.e. owned or operated by different Cloud operators, to maximize the 132 benefits of geographical proximity, elasticity and special features 133 offered by different cloud DCs. 135 1.2. The role of SD-WAN techniques in Cloud DC connectivity 137 This document discusses the issues associated with connecting 138 enterprise to their workloads/applications instantiated in multiple 139 third-party data centers (a.k.a. Cloud DCs). Very often, the actual 140 Cloud DCs that host the workloads/applications can be transient. . 142 SD-WAN, initially launched to maximize bandwidths between locations 143 by aggregating multiple paths managed by different service 144 providers, has expanded to include flexible, on-demand, application- 145 based connections established over any networks to access dynamic 146 workloads in Cloud DCs. 148 As a consequence, this document discusses the use of SD-WAN 149 techniques as a means to improve enterprise-to-cloud DC 150 connectivity. 152 2. Definition of terms 154 Cloud DC: Third party Data Centers that usually host applications 155 and workload owned by different organizations or 156 tenants. 158 Controller: Used interchangeably with SD-WAN controller to manage 159 SD-WAN overlay path creation/deletion and monitoring the 160 path conditions between two or more sites. 162 DSVPN: Dynamic Smart Virtual Private Network. DSVPN is a secure 163 network that exchanges data between sites without 164 needing to pass traffic through an organization's 165 headquarter virtual private network (VPN) server or 166 router. 168 Heterogeneous Cloud: applications and workloads split among Cloud 169 DCs owned or managed by different operators. 171 Hybrid Clouds: Hybrid Clouds refers to an enterprise using its own 172 on-premises DCs in addition to Cloud services provided 173 by one or more cloud operators. (e.g. AWS, Azure, 174 Google, Salesforces, SAP, etc). 176 SD-WAN: Software Defined Wide Area Network. In this document, 177 "SD-WAN" refers to the solutions of pooling WAN 178 bandwidth from multiple underlay networks to get better 179 WAN bandwidth management, visibility & control. When the 180 underlay networks are private networks, traffic can 181 traverse without additional encryption; when the 182 underlay networks are public, such as Internet, some 183 traffic needs to be encrypted when traversing through 184 (depending on user provided policies). 186 VPC: Virtual Private Cloud. A service offered by Cloud DC 187 operators to allocate logically-isolated cloud 188 resources, including compute, networking and storage. 190 3. Current Practices in Interconnecting Enterprise Sites with Cloud DCs 192 3.1. Multiple connection to workloads in a Cloud DC 194 Most Cloud operators offer some type of network gateway through 195 which an enterprise can reach their workloads hosted in the Cloud 196 DCs. For example, AWS (Amazon Web Services) offers the following 197 options to reach workloads in AWS Cloud DCs: 199 - AWS Internet gateway allows communication between instances in 200 AWS VPC and the internet. 201 - AWS Virtual gateway (vGW) where IPsec tunnels [RFC6071] are 202 established between an enterprise's own gateway and AWS vGW, so 203 that the communications between those gateways can be secured 204 from the underlay (which might be the public Internet). 205 - AWS Direct Connect, which allows enterprises to purchase direct 206 connect from network service providers to get a private leased 207 line interconnecting the enterprises gateway(s) and the AWS 208 Direct Connect routers. In addition, an AWS Transit Gateway can 209 be used to interconnect multiple VPCs in different Availability 210 Zones. 212 As an example, some branch offices of an enterprise can connect to 213 over the Internet to reach AWS's vGW via IPsec tunnels. Other branch 214 offices of the same enterprise can connect to AWS DirectConnect via 215 a private network (without any encryption). ). It is important for 216 enterprises to be able to observe the specific behaviors when 217 connected by different connections. 219 Figure below shows an example of some tenants' workloads are 220 accessible via a virtual router connected by AWS Internet Gateway; 221 some are accessible via AWS vGW, and others are accessible via AWS 222 Direct Connect. The vR1 can have its own IPsec capability for secure 223 tunnel over the internet to bypass paying additional price for the 224 IPsec features provided by AWS vGW. Some tenants can deploy separate 225 virtual routers to connect to internet traffic and to traffic from 226 the secure channels from vGW and DirectConnect, e.g. vR1 & vR2. 227 Others may have one virtual router connecting to both types of 228 traffic. Customer Gateway can be customer owned router or ports 229 physically connected to AWS Direct Connect GW. 231 +------------------------+ 232 | ,---. ,---. | 233 | (TN-1 ) ( TN-2)| 234 | `-+-' +---+ `-+-' | 235 | +----|vR1|----+ | 236 | ++--+ | 237 | | +-+----+ 238 | | /Internet\ For External 239 | +-------+ Gateway +---------------------- 240 | \ / to reach via Internet 241 | +-+----+ 242 | | 243 | ,---. ,---. | 244 | (TN-1 ) ( TN-2)| 245 | `-+-' +---+ `-+-' | 246 | +----|vR2|----+ | 247 | ++--+ | 248 | | +-+----+ 249 | | / virtual\ For IPsec Tunnel 250 | +-------+ Gateway +---------------------- 251 | | \ / termination 252 | | +-+----+ 253 | | | 254 | | +-+----+ +------+ 255 | | / \ For Direct /customer\ 256 | +-------+ Gateway +----------+ gateway | 257 | \ / Connect \ / 258 | +-+----+ +------+ 259 | | 260 +------------------------+ 262 Figure 1: Examples of Multiple Cloud DC connections. 264 3.2. Interconnect to Hybrid Cloud DCs 266 Hybrid will be the most common usage of the cloud as more 267 enterprises see the benefits of integrating public and private cloud 268 infrastructures. However, enabling the growth of hybrid cloud 269 deployments in the enterprise requires fast and safe interconnection 270 between public and private cloud services. 271 For an enterprise to connect to applications & workloads hosted in 272 multiple Cloud DCs, the enterprise can use IPsec tunnels established 273 over the Internet or a (virtualized) leased line service to connect 274 its on-premises gateways to each of the Cloud DC's gateways, virtual 275 routers instantiated in the Cloud DCs, or any other suitable design 276 (including a combination thereof). 278 Some enterprises prefer to instantiate their own virtual 279 CPEs/routers inside the Cloud DC to connect the workloads within the 280 Cloud DC. Then an overlay path is established between customer 281 gateways to the virtual CPEs/routers for reaching the workloads 282 inside the cloud DC. 284 3.3. Connecting workloads among hybrid Cloud DCs 286 There are multiple approaches to interconnect workloads among 287 different Cloud DCs: 289 a) Utilize Cloud DC provided transit gateways. 290 b) Hairpin all the traffic through the customer gateway, or 291 c) Establish direct tunnels among different VPCs (Virtual Private 292 Clouds) via client's own virtual routers instantiated within 293 Cloud DCs. DMVPN (Dynamic Multipoint Virtual Private Network) 294 or DSVPN (Dynamic Smart VPN) techniques can be used to 295 establish direct Multi-point-to-Point or multi-point-to multi- 296 point tunnels among those client's own virtual routers. 298 Approach a) usually does not work if Cloud DCs are owned and managed 299 by different Cloud providers. 301 Approach b) creates additional transmission delay plus incurring 302 cost when exiting Cloud DCs. 304 For the Approach c), DMVPN or DSVPN use NHRP (Next Hop Resolution 305 Protocol) [RFC2735] so that spoke nodes can register their IP 306 addresses & WAN ports with the hub node. The IETF ION 307 (Internetworking over NBMA (non-broadcast multiple access) WG 308 standardized NHRP for connection-oriented NBMA network (such as ATM) 309 network address resolution more than two decades ago. 311 There are many differences between virtual routers in Public Cloud 312 DCs and the nodes in an NBMA network. NHRP cannot be used for 313 registering virtual routers in Cloud DCs unless an extension of such 314 protocols is developed for that purpose. Therefore, DMVPN and/or 315 DSVPN cannot be used directly for connecting workloads in hybrid 316 Cloud DCs. 318 Other protocols such as BGP can be used, as described in [BGP- 319 SDWAN]. 321 4. Desired Properties for Networks that interconnect Hybrid Clouds 322 The networks that interconnect hybrid cloud DCs must address the 323 following requirements: 324 - High availability to access all workloads in the desired cloud 325 DCs. 326 Many enterprises include cloud infrastructures in their 327 disaster recovery strategy, e.g., by enforcing periodic backup 328 policies within the cloud, or by running backup applications in 329 the Cloud, etc. Therefore, the connection to the cloud DCs may 330 not be permanent, but rather needs to be on-demand. 332 - Global reachability from different geographical zones, thereby 333 facilitating the proximity of applications as a function of the 334 end users' location, to improve latency. 335 - Elasticity: prompt connection to newly instantiated 336 applications at Cloud DCs when end-users' usages increase and 337 prompt release of connection after applications at locations 338 being removed when demands change. 339 Some enterprises have front-end web portals running in cloud 340 DCs and database servers in their on-premises DCs. Those Front- 341 end web portals need to be reachable from the public Internet. 342 The backend connection to the sensitive data in database 343 servers hosted in the on-premises DCs might need secure 344 connections. 346 - Scalable security management. IPsec is commonly used to 347 interconnect cloud gateways with CPEs deployed in the 348 enterprise premises. For enterprises with a large number or 349 branch offices, managing the IPsec's Security Associations 350 among many nodes can be very difficult. 352 5. Problems with MPLS-based VPNs extending to Hybrid Cloud DCs 354 Traditional MPLS-based VPNs have been widely deployed as an 355 effective way to support businesses and organizations that require 356 network performance and reliability. MPLS shifted the burden of 357 managing a VPN service from enterprises to service providers. The 358 CPEs attached to MPLS VPNs are also simpler and less expensive, 359 since they do not need to manage routes to remote sites; they simply 360 pass all outbound traffic to the MPLS VPN PEs to which the CPEs are 361 attached (albeit multi-homing scenarios require more processing 362 logic on CPEs). MPLS has addressed the problems of scale, 363 availability, and fast recovery from network faults, and 364 incorporated traffic-engineering capabilities. 366 However, traditional MPLS-based VPN solutions are sub-optimized for 367 connecting end-users to dynamic workloads/applications in cloud DCs 368 because: 370 - The Provider Edge (PE) nodes of the enterprise's VPNs might not 371 have direct connections to third party cloud DCs that are used 372 for hosting workloads with the goal of providing an easy access 373 to enterprises' end-users. 375 - It usually takes some time to deploy provider edge (PE) routers 376 at new locations. When enterprise's workloads are changed from 377 one cloud DC to another (i.e., removed from one DC and re- 378 instantiated to another location when demand changes), the 379 enterprise branch offices need to be connected to the new cloud 380 DC, but the network service provider might not have PEs located 381 at the new location. 383 One of the main drivers for moving workloads into the cloud is 384 the widely available cloud DCs at geographically diverse 385 locations, where apps can be instantiated so that they can be 386 as close to their end-users as possible. When the user base 387 changes, the applications may be migrated to a new cloud DC 388 location closest to the new user base. 390 - Most of the cloud DCs do not expose their internal networks, so 391 the MPLS-based VPNs can only reach Cloud DC's Gateways, not to 392 the workloads hosted inside. Even with AWS DirectConnect, the 393 connection only reaches the AWS DirectConnect Gateway. 395 - Extensive usage of Overlay by Cloud DCs: 397 Many cloud DCs use an overlay to connect their gateways to the 398 workloads located inside the DC. There is currently no standard 399 that specifies the interworking between the Cloud Overlay and the 400 enterprise' existing underlay networks. One of the 401 characteristics of overlay networks is that some of the WAN ports rd of the edge nodes connect to 3 party networks. There is 402 therefore a need to propagate WAN port information to remote rd authorized peers in 3 party network domains in addition to route 403 propagation. Such an exchange cannot happen before communication 404 between peers is properly secured. 406 Another roadblock is the lack of a standard way to express and 407 enforce consistent security policies for workloads that not only use 408 virtual addresses, but in which are also very likely hosted in 409 different locations within the Cloud DC [RFC8192]. The current VPN 410 path computation and bandwidth allocation schemes may not be 411 flexible enough to address the need for enterprises to rapidly 412 connect to dynamically instantiated (or removed) workloads and 413 applications regardless of their location/nature (i.e., third party 414 cloud DCs). 416 6. Problem with using IPsec tunnels to Cloud DCs 417 As described in the previous section, many Cloud operators expose 418 their gateways for external entities (which can be enterprises 419 themselves) to directly establish IPsec tunnels. Enterprises can 420 also instantiate virtual routers within Cloud DCs to connect to 421 their on-premises devices via IPsec tunnels. If there is only one 422 enterprise location that needs to reach the Cloud DC, an IPsec 423 tunnel is a very convenient solution. 425 However, many medium-to-large enterprises usually have multiple 426 sites and multiple data centers. For workloads and apps hosted in 427 cloud DCs, multiple sites need to communicate securely with those 428 cloud workloads and apps. This section documents some of the issues 429 associated with using IPsec tunnels to connect enterprise premises 430 with cloud gateways. 432 6.1. Complexity of multi-point any-to-any interconnection 434 The dynamic workload instantiated in cloud DC needs to communicate 435 with multiple branch offices and on-premises data centers. Most 436 enterprises need multi-point interconnection among multiple 437 locations, which can be provided by means of MPLS L2/L3 VPNs. 439 Using IPsec overlay paths to connect all branches & on-premises data 440 centers to cloud DCs requires CPEs to manage routing among Cloud DCs 441 gateways and the CPEs located at other branch locations, which can 442 dramatically increase the complexity of the design, possibly at the 443 cost of jeopardizing the CPE performance. 445 The complexity of requiring CPEs to maintain routing among other 446 CPEs is one of the reasons why enterprises migrated from Frame Relay 447 based services to MPLS-based VPN services. 449 MPLS-based VPNs have their PEs directly connected to the CPEs. 450 Therefore, CPEs only need to forward all traffic to the directly 451 attached PEs, which are therefore responsible for enforcing the 452 routing policy within the corresponding VPNs. Even for multi-homed 453 CPEs, the CPEs only need to forward traffic among the directly 454 connected PEs. However, when using IPsec tunnels between CPEs and 455 Cloud DCs, the CPEs need to compute, select, establish and maintain 456 routes for traffic to be forwarded to Cloud DCs, to remote CPEs via 457 VPN, or directly. 459 6.2. Poor performance over long distance 461 When enterprise CPEs or gateways are far away from cloud DC gateways 462 or across country/continent boundaries, performance of IPsec tunnels 463 over the public Internet can be problematic and unpredictable. Even 464 though there are many monitoring tools available to measure delay 465 and various performance characteristics of the network, the 466 measurement for paths over the Internet is passive and past 467 measurements may not represent future performance. 469 Many cloud providers can replicate workloads in different available 470 zones. An App instantiated in a cloud DC closest to clients may have 471 to cooperate with another App (or its mirror image) in another 472 region or database server(s) in the on-premises DC. This kind of 473 coordination requires predicable networking behavior/performance 474 among those locations. 476 6.3. Scaling Issues with IPsec Tunnels 478 IPsec can achieve secure overlay connections between two locations 479 over any underlay network, e.g., between CPEs and Cloud DC Gateways. 481 If there is only one enterprise location connected to the cloud 482 gateway, a small number of IPsec tunnels can be configured on-demand 483 between the on-premises DC and the Cloud DC, which is an easy and 484 flexible solution. 486 However, for multiple enterprise locations to reach workloads hosted 487 in cloud DCs, the cloud DC gateway needs to maintain multiple IPsec 488 tunnels to all those locations (e.g., as a hub & spoke topology). 489 For a company with hundreds or thousands of locations, there could 490 be hundreds (or even thousands) of IPsec tunnels terminating at the 491 cloud DC gateway, which is not only very expensive (because Cloud 492 Operators usually charge their customers based on connections), but 493 can be very processing intensive for the gateway. Many cloud 494 operators only allow a limited number of (IPsec) tunnels & bandwidth 495 to each customer. Alternatively, you could use a solution like 496 group encryption where a single IPsec SA is necessary at the GW but 497 the drawback here is key distribution and maintenance of a key 498 server, etc. 500 7. Problems of Using SD-WAN to connect to Cloud DCs 501 SD-WAN can establish parallel paths over multiple underlay networks 502 between two locations on-demand, for example, to support the 503 connections established between two CPEs interconnected by a 504 traditional MPLS VPN ([RFC4364] or [RFC4664]) or by IPsec [RFC6071] 505 tunnels. 507 SD-WAN lets enterprises augment their current VPN network with cost- 508 effective, readily available Broadband Internet connectivity, 509 enabling some traffic offloading to paths over the Internet 510 according to differentiated, possibly application-based traffic 511 forwarding policies, or when the MPLS VPN connection between the two 512 locations is congested, or otherwise undesirable or unavailable. 514 7.1. SD-WAN among branch offices vs. interconnect to Cloud DCs 516 SD-WAN interconnection of branch offices is not as simple as it 517 appears. For an enterprise with multiple sites, using SD-WAN overlay 518 paths among sites requires each CPE to manage all the addresses that 519 local hosts have the potential to reach, i.e., map internal VPN 520 addresses to appropriate SD-WAN paths. This is similar to the 521 complexity of Frame Relay based VPNs, where each CPE needed to 522 maintain mesh routing for all destinations if they were to avoid an 523 extra hop through a hub router. Even though SD-WAN CPEs can get 524 assistance from a central controller (instead of running a routing 525 protocol) to resolve the mapping between destinations and SD-WAN 526 paths, SD-WAN CPEs are still responsible for routing table 527 maintenance as remote destinations change their attachments, e.g., 528 the dynamic workload in other DCs are de-commissioned or added. 530 Even though originally envisioned for interconnecting branch 531 offices, SD-WAN offers a very attractive way for enterprises to 532 connect to Cloud DCs. 534 The SD-WAN for interconnecting branch offices and the SD-WAN for 535 interconnecting to Cloud DCs have some differences: 537 - SD-WAN for interconnecting branch offices usually have two end- 538 points (e.g., CPEs) controlled by one entity (e.g., a 539 controller or management system operated by the enterprise). 540 - SD-WAN for Cloud DC interconnects may consider CPEs owned or 541 managed by the enterprise, while remote end-points are being 542 managed or controlled by Cloud DCs (For the ease of 543 description, let's call such CPEs asymmetrically-managed CPEs). 545 - Cloud DCs may have different entry points (or devices) with one 546 entry point that terminates a private direct connection (based 547 upon a leased line for example) and other entry points being 548 devices terminating the IPsec tunnels, as shown in Figure 2. 550 Therefore, the SD-WAN design becomes asymmetric. 551 +------------------------+ 552 | ,---. ,---. | 553 | (TN-1 ) ( TN-2)| TN: Tenant applications/workloads 554 | `-+-' +---+ `-+-' | 555 | +----|vR1|----+ | 556 | ++--+ | 557 | | +-+----+ 558 | | /Internet\ One path via 559 | +-------+ Gateway +---------------------+ 560 | \ / Internet \ 561 | +-+----+ \ 562 +------------------------+ \ 563 \ 564 +------------------------+ native traffic \ 565 | ,---. ,---. | without encryption| 566 | (TN-3 ) ( TN-4)| | 567 | `-+-' +--+ `-+-' | | +------+ 568 | +----|vR|-----+ | +--+ CPE | 569 | ++-+ | | +------+ 570 | | +-+----+ | 571 | | / virtual\ One path via IPsec Tunnel | 572 | +-------+ Gateway +-------------------------- + 573 | \ / Encrypted traffic over| 574 | +-+----+ public network | 575 +------------------------+ | 576 | 577 +------------------------+ | 578 | ,---. ,---. | Native traffic | 579 | (TN-5 ) ( TN-6)| without encryption | 580 | `-+-' +--+ `-+-' | over secure network| 581 | +----|vR|-----+ | | 582 | ++-+ | | 583 | | +-+----+ +------+ | 584 | | / \ Via Direct /customer\ | 585 | +-------+ Gateway +----------+ gateway |-----+ 586 | \ / Connect \ / 587 | +-+----+ +------+ 588 +------------------------+ Customer GW has physical connection to AWS GW. 590 Figure 2: Different Underlays to Reach Cloud DC 592 8. End-to-End Security Concerns for Data Flows 594 When IPsec tunnels established from enterprise on-premises CPEs 595 are terminated at the Cloud DC gateway where the workloads or 596 applications are hosted, some enterprises have concerns regarding 597 traffic to/from their workload being exposed to others behind the 598 data center gateway (e.g., exposed to other organizations that 599 have workloads in the same data center). 600 To ensure that traffic to/from workloads is not exposed to 601 unwanted entities, IPsec tunnels may go all the way to the 602 workload (servers, or VMs) within the DC. 604 9. Requirements for Dynamic Cloud Data Center VPNs 606 In order to address the aforementioned issues, any solution for 607 enterprise VPNs that includes connectivity to dynamic workloads or 608 applications in cloud data centers should satisfy a set of 609 requirements: 611 - The solution should allow enterprises to take advantage of the 612 current state-of-the-art in VPN technology, in both traditional 613 MPLS-based VPNs and IPsec-based VPNs (or any combination 614 thereof) that run over the public Internet. 615 - The solution should not require an enterprise to upgrade all 616 their existing CPEs. 617 - The solution should support scalable IPsec key management among 618 all nodes involved in DC interconnect schemes. 619 - The solution needs to support easy and fast, on-the-fly, VPN 620 connections to dynamic workloads and applications in third 621 party data centers, and easily allow these workloads to migrate 622 both within a data center and between data centers. 623 - Allow VPNs to provide bandwidth and other performance 624 guarantees. 625 - Be a cost-effective solution for enterprises to incorporate 626 dynamic cloud-based applications and workloads into their 627 existing VPN environment. 629 10. Security Considerations 631 The draft discusses security requirements as a part of the problem 632 space, particularly in sections 4, 5, and 8. 634 Solution drafts resulting from this work will address security 635 concerns inherent to the solution(s), including both protocol 636 aspects and the importance (for example) of securing workloads in 637 cloud DCs and the use of secure interconnection mechanisms. 639 11. IANA Considerations 641 This document requires no IANA actions. RFC Editor: Please remove 642 this section before publication. 644 12. References 646 12.1. Normative References 648 12.2. Informative References 650 [RFC2735] B. Fox, et al "NHRP Support for Virtual Private 651 networks". Dec. 1999. 653 [RFC8192] S. Hares, et al "Interface to Network Security Functions 654 (I2NSF) Problem Statement and Use Cases", July 2017 656 [ITU-T-X1036] ITU-T Recommendation X.1036, "Framework for creation, 657 storage, distribution and enforcement of policies for 658 network security", Nov 2007. 660 [RFC6071] S. Frankel and S. Krishnan, "IP Security (IPsec) and 661 Internet Key Exchange (IKE) Document Roadmap", Feb 2011. 663 [RFC4364] E. Rosen and Y. Rekhter, "BGP/MPLS IP Virtual Private 664 Networks (VPNs)", Feb 2006 666 [RFC4664] L. Andersson and E. Rosen, "Framework for Layer 2 Virtual 667 Private Networks (L2VPNs)", Sept 2006. 669 [BGP-SDWAN] L. Dunbar, et al. "BGP Extension for SDWAN Overlay 670 Networks", draft-dunbar-idr-bgp-sdwan-overlay-ext-03, 671 work-in-progress, Nov 2018. 673 13. Acknowledgments 675 Many thanks to Chris Bowers, Ignas Bagdonas, Michael Huang, Liu Yuan 676 Jiao, Katherine Zhao, and Jim Guichard for the discussion and 677 contributions. 679 Authors' Addresses 681 Linda Dunbar 682 Futurewei 683 Email: Linda.Dunbar@futurewei.com 685 Andrew G. Malis 686 Futurewei 687 Email: agmalis@gmail.com 689 Christian Jacquenet 690 Orange 691 Rennes, 35000 692 France 693 Email: Christian.jacquenet@orange.com 695 Mehmet Toy 696 Verizon 697 One Verizon Way 698 Basking Ridge, NJ 07920 699 Email: mehmet.toy@verizon.com