idnits 2.17.00 (12 Aug 2021) /tmp/idnits61401/draft-ietf-rtgwg-net2cloud-problem-statement-04.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 8 instances of too long lines in the document, the longest one being 85 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 (September 23, 2019) is 971 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Unused Reference: 'ITU-T-X1036' is defined on line 762, 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 Futurewei 3 Intended status: Informational Andy Malis 4 Expires: Dec 2019 Independent 5 C. Jacquenet 6 Orange 7 M. Toy 8 Verizon 9 September 23, 2019 11 Dynamic Networks to Hybrid Cloud DCs Problem Statement 12 draft-ietf-rtgwg-net2cloud-problem-statement-04 14 Abstract 16 This document describes the problems that enterprises face today 17 when interconnecting their branch offices with dynamic workloads in 18 third party data centers (a.k.a. Cloud DCs). 20 It examines some of the approaches interconnecting cloud DCs with 21 enterprises' on-premises DCs & branch offices. This document also 22 describes some of the network problems that many enterprises face 23 when they have workloads & applications & data split among different 24 data centers, especially for those enterprises with multiple sites 25 that are already interconnected by VPNs (e.g., MPLS L2VPN/L3VPN). 27 Current operational problems are examined to determine whether there 28 is a need to improve existing protocols or whether a new protocol is 29 necessary to solve them. 31 Status of this Memo 33 This Internet-Draft is submitted in full conformance with the 34 provisions of BCP 78 and BCP 79. 36 Internet-Drafts are working documents of the Internet Engineering 37 Task Force (IETF), its areas, and its working groups. Note that 38 other groups may also distribute working documents as Internet- 39 Drafts. 41 Internet-Drafts are draft documents valid for a maximum of six 42 months and may be updated, replaced, or obsoleted by other documents 43 at any time. It is inappropriate to use Internet-Drafts as 44 reference material or to cite them other than as "work in progress." 46 The list of current Internet-Drafts can be accessed at 47 http://www.ietf.org/ietf/1id-abstracts.txt 49 The list of Internet-Draft Shadow Directories can be accessed at 50 http://www.ietf.org/shadow.html 52 This Internet-Draft will expire on March 23, 2009. 54 Copyright Notice 56 Copyright (c) 2019 IETF Trust and the persons identified as the 57 document authors. All rights reserved. 59 This document is subject to BCP 78 and the IETF Trust's Legal 60 Provisions Relating to IETF Documents 61 (http://trustee.ietf.org/license-info) in effect on the date of 62 publication of this document. Please review these documents 63 carefully, as they describe your rights and restrictions with 64 respect to this document. Code Components extracted from this 65 document must include Simplified BSD License text as described in 66 Section 4.e of the Trust Legal Provisions and are provided without 67 warranty as described in the Simplified BSD License. 69 Table of Contents 71 1. Introduction...................................................3 72 1.1. On the evolution of Cloud DC connectivity.................3 73 1.2. The role of SD-WAN techniques in Cloud DC connectivity....4 74 2. Definition of terms............................................4 75 3. Interconnecting Enterprise Sites with Cloud DCs................5 76 3.1. Multiple connections to workloads in a Cloud DC...........5 77 3.2. Interconnect Private and Public Cloud DCs.................7 78 3.3. Desired Properties for Networks that interconnect Hybrid 79 Clouds.........................................................8 80 4. Multiple Clouds Interconnection................................9 81 4.1. Multi-Cloud Interconnection...............................9 82 4.2. Desired Properties for Multi-Cloud Interconnection.......11 84 5. Problems with MPLS-based VPNs extending to Hybrid Cloud DCs...11 85 6. Problem with using IPsec tunnels to Cloud DCs.................13 86 6.1. Complexity of multi-point any-to-any interconnection.....13 87 6.2. Poor performance over long distance......................14 88 6.3. Scaling Issues with IPsec Tunnels........................14 89 7. Problems of Using SD-WAN to connect to Cloud DCs..............15 90 7.1. SD-WAN among branch offices vs. interconnect to Cloud DCs15 91 8. End-to-End Security Concerns for Data Flows...................18 92 9. Requirements for Dynamic Cloud Data Center VPNs...............18 93 10. Security Considerations......................................19 94 11. IANA Considerations..........................................19 95 12. References...................................................19 96 12.1. Normative References....................................19 97 12.2. Informative References..................................19 98 13. Acknowledgments..............................................20 100 1. Introduction 102 1.1. On the evolution of Cloud DC connectivity 104 The ever-increasing use of cloud applications for communication 105 services change the way corporate business works and shares 106 information. Such cloud applications use resources hosted in third 107 party DCs that also host services for other customers. 109 With the advent of widely available third-party cloud DCs in diverse 110 geographic locations and the advancement of tools for monitoring and 111 predicting application behaviors, it is technically feasible for 112 enterprises to instantiate applications and workloads in locations 113 that are geographically closest to their end-users. Such proximity 114 improves end-to-end latency and overall user experience. Conversely, 115 an enterprise can easily shutdown applications and workloads 116 whenever end-users are in motion (thereby modifying the networking 117 connection of subsequently relocated applications and workloads). In 118 addition, an enterprise may wish to take advantage of more and more 119 business applications offered by third party private cloud DCs. 121 Most of those enterprise branch offices & on-premises data centers 122 are already connected via VPNs, such as MPLS-based L2VPNs and 123 L3VPNs. Then connecting to the cloud-hosted resources may not be 124 straightforward if the provider of the VPN service does not have 125 direct connections to the corresponding cloud DCs. Under those 126 circumstances, the enterprise can upgrade the CPEs deployed in its 127 various premises to utilize SD-WAN techniques to reach cloud 128 resources (without any assistance from the VPN service provider), or 129 wait for their VPN service provider to make new agreements with data 130 center providers to connect to the cloud resources. Either way has 131 additional infrastructure and operational costs. 133 In addition, more enterprises are moving towards hybrid cloud DCs, 134 i.e. owned or operated by different Cloud operators, to maximize the 135 benefits of geographical proximity, elasticity and special features 136 offered by different cloud DCs. 138 1.2. The role of SD-WAN techniques in Cloud DC connectivity 140 This document discusses the issues associated with connecting 141 enterprise's workloads/applications instantiated in multiple third- 142 party data centers (a.k.a. Cloud DCs) and its on-prem data centers. 143 Very often, the actual Cloud DCs that host the 144 workloads/applications can be transient. 146 SD-WAN, initially launched to maximize bandwidths between locations 147 by aggregating multiple paths managed by different service 148 providers, has expanded to include flexible, on-demand, application- 149 based connections established over any networks to access dynamic 150 workloads in Cloud DCs. 152 Therefore, this document discusses the use of SD-WAN techniques to 153 improve enterprise-to-cloud DC and cloud DC-to-cloud DC 154 connectivity. 156 2. Definition of terms 158 Cloud DC: Third party Data Centers that usually host applications 159 and workload owned by different organizations or 160 tenants. 162 Controller: Used interchangeably with SD-WAN controller to manage 163 SD-WAN overlay path creation/deletion and monitoring the 164 path conditions between two or more sites. 166 DSVPN: Dynamic Smart Virtual Private Network. DSVPN is a secure 167 network that exchanges data between sites without 168 needing to pass traffic through an organization's 169 headquarter virtual private network (VPN) server or 170 router. 172 Heterogeneous Cloud: applications and workloads split among Cloud 173 DCs owned or managed by different operators. 175 Hybrid Clouds: Hybrid Clouds refers to an enterprise using its own 176 on-premises DCs in addition to Cloud services provided 177 by one or more cloud operators. (e.g. AWS, Azure, 178 Google, Salesforces, SAP, etc). 180 SD-WAN: Software Defined Wide Area Network. In this document, 181 "SD-WAN" refers to the solutions of pooling WAN 182 bandwidth from multiple underlay networks to get better 183 WAN bandwidth management, visibility & control. When the 184 underlay networks are private networks, traffic can 185 traverse without additional encryption; when the 186 underlay networks are public, such as Internet, some 187 traffic needs to be encrypted when traversing through 188 (depending on user provided policies). 190 VPC: Virtual Private Cloud is a virtual network dedicated to 191 one client account. It is logically isolated from other 192 virtual networks in a Cloud DC. Each client can launch 193 his/her desired resources, such as compute, storage, or 194 network functions into his/her VPC. Most Cloud 195 operators' VPCs only support private addresses, some 196 support IPv4 only, others support IPv4/IPv6 dual stack. 198 3. Interconnecting Enterprise Sites with Cloud DCs 200 3.1. Multiple connections to workloads in a Cloud DC 202 Most Cloud operators offer some type of network gateway through 203 which an enterprise can reach their workloads hosted in the Cloud 204 DCs. For example, AWS (Amazon Web Services) offers the following 205 options to reach workloads in AWS Cloud DCs: 207 - AWS Internet gateway allows communication between instances in 208 AWS VPC and the internet. 210 - AWS Virtual gateway (vGW) where IPsec tunnels [RFC6071] are 211 established between an enterprise's own gateway and AWS vGW, so 212 that the communications between those gateways can be secured 213 from the underlay (which might be the public Internet). 214 - AWS Direct Connect, which allows enterprises to purchase direct 215 connect from network service providers to get a private leased 216 line interconnecting the enterprises gateway(s) and the AWS 217 Direct Connect routers. In addition, an AWS Transit Gateway can 218 be used to interconnect multiple VPCs in different Availability 219 Zones. AWS Transit Gateway acts as a hub that controls how 220 traffic is forwarded among all the connected networks which act 221 like spokes. 223 As an example, some branch offices of an enterprise can connect to 224 over the Internet to reach AWS's vGW via IPsec tunnels. Other branch 225 offices of the same enterprise can connect to AWS DirectConnect via 226 a private network (without any encryption). ). It is important for 227 enterprises to be able to observe the specific behaviors when 228 connected by different connections. 230 Figure below shows an example of some tenants' workloads are 231 accessible via a virtual router connected by AWS Internet Gateway; 232 some are accessible via AWS vGW, and others are accessible via AWS 233 Direct Connect. vR1 uses IPsec to establish secure tunnels over the 234 Internet to avoid paying extra fees for the IPsec features provided 235 by AWS vGW. Some tenants can deploy separate virtual routers to 236 connect to internet traffic and to traffic from the secure channels 237 from vGW and DirectConnect, e.g. vR1 & vR2. Others may have one 238 virtual router connecting to both types of traffic. Customer Gateway 239 can be customer owned router or ports physically connected to AWS 240 Direct Connect GW. 242 +------------------------+ 243 | ,---. ,---. | 244 | (TN-1 ) ( TN-2)| 245 | `-+-' +---+ `-+-' | 246 | +----|vR1|----+ | 247 | ++--+ | 248 | | +-+----+ 249 | | /Internet\ For External 250 | +-------+ Gateway +---------------------- 251 | \ / to reach via Internet 252 | +-+----+ 253 | | 254 | ,---. ,---. | 255 | (TN-1 ) ( TN-2)| 256 | `-+-' +---+ `-+-' | 257 | +----|vR2|----+ | 258 | ++--+ | 259 | | +-+----+ 260 | | / virtual\ For IPsec Tunnel 261 | +-------+ Gateway +---------------------- 262 | | \ / termination 263 | | +-+----+ 264 | | | 265 | | +-+----+ +------+ 266 | | / \ For Direct /customer\ 267 | +-------+ Gateway +----------+ gateway | 268 | \ / Connect \ / 269 | +-+----+ +------+ 270 | | 271 +------------------------+ 273 Figure 1: Examples of Multiple Cloud DC connections. 275 3.2. Interconnect Private and Public Cloud DCs 277 It is likely that hybrid designs will become the rule for cloud 278 services, as more enterprises see the benefits of integrating public 279 and private cloud infrastructures. However, enabling the growth of 280 hybrid cloud deployments in the enterprise requires fast and safe 281 interconnection between public and private cloud services. 282 For an enterprise to connect to applications & workloads hosted in 283 multiple Cloud DCs, the enterprise can use IPsec tunnels established 284 over the Internet or a (virtualized) leased line service to connect 285 its on-premises gateways to each of the Cloud DC's gateways, virtual 286 routers instantiated in the Cloud DCs, or any other suitable design 287 (including a combination thereof). 289 Some enterprises prefer to instantiate their own virtual 290 CPEs/routers inside the Cloud DC to connect the workloads within the 291 Cloud DC. Then an overlay path is established between customer 292 gateways to the virtual CPEs/routers for reaching the workloads 293 inside the cloud DC. 295 3.3. Desired Properties for Networks that interconnect Hybrid Clouds 297 The networks that interconnect hybrid cloud DCs must address the 298 following requirements: 299 - High availability to access all workloads in the desired cloud 300 DCs. 301 Many enterprises include cloud infrastructures in their 302 disaster recovery strategy, e.g., by enforcing periodic backup 303 policies within the cloud, or by running backup applications in 304 the Cloud, etc. Therefore, the connection to the cloud DCs may 305 not be permanent, but rather needs to be on-demand. 307 - Global reachability from different geographical zones, thereby 308 facilitating the proximity of applications as a function of the 309 end users' location, to improve latency. 310 - Elasticity: prompt connection to newly instantiated 311 applications at Cloud DCs when usages increase and prompt 312 release of connection after applications at locations being 313 removed when demands change. 314 Some enterprises have front-end web portals running in cloud 315 DCs and database servers in their on-premises DCs. Those Front- 316 end web portals need to be reachable from the public Internet. 317 The backend connection to the sensitive data in database 318 servers hosted in the on-premises DCs might need secure 319 connections. 321 - Scalable security management. IPsec is commonly used to 322 interconnect cloud gateways with CPEs deployed in the 323 enterprise premises. For enterprises with a large number or 324 branch offices, managing the IPsec's Security Associations 325 among many nodes can be very difficult. 327 4. Multiple Clouds Interconnection 329 4.1. Multi-Cloud Interconnection 331 Enterprises today can instantiate their workloads or applications in 332 Cloud DCs owned by different Cloud providers, e.g. AWS, Azure, 333 GoogleCloud, Oracle, etc. Interconnecting those workloads involves 334 three parties: The Enterprise, its network service providers, and 335 the Cloud providers. 337 All Cloud Operators offer secure ways to connect enterprises' on- 338 prem sites/DCs with their Cloud DCs. 340 Some Cloud Operators allow enterprises to connect via private 341 networks. For example, AWS's DirectConnect allows enterprises to use rd 3 party provided private Layer 2 path from enterprises' GW to AWS 342 DirectConnect GW. Microsoft's ExpressRoute allows extension of a 343 private network to any of the Microsoft cloud services, including 344 Azure and Office365. ExpressRoute is configured using Layer 3 345 routing. Customers can opt for redundancy by provisioning dual links 346 from their location to two Microsoft Enterprise edge routers (MSEEs) 347 located within a third-party ExpressRoute peering location. The BGP 348 routing protocol is then setup over WAN links to provide redundancy 349 to the cloud. This redundancy is maintained from the peering data 350 center into Microsoft's cloud network. 352 Google's Cloud Dedicated Interconnect offers similar network 353 connectivity options as AWS and Microsoft. One distinct difference, 354 however, is that Google's service allows customers access to the 355 entire global cloud network by default. It does this by connecting 356 your on-premises network with the Google Cloud using BGP and Google 357 Cloud Routers to provide optimal paths to the different regions of 358 the global cloud infrastructure. 360 All those connectivity options are between Cloud providers' DCs and 361 the Enterprises, but not between cloud DCs. For example, to connect 362 applications in AWS Cloud to applications in Azure Cloud, there must 363 be a third-party gateway (physical or virtual) to interconnect the 364 AWS's Layer 2 DirectConnect path with Azure's Layer 3 ExpressRoute. 366 Enterprises can also instantiate their own virtual routers in 367 different Cloud DCs and administer IPsec tunnels among them, which 368 by itself is not a trivial task. Or by leveraging open source VPN 369 software such as strongSwan, you create an IPSec connection to the 370 Azure gateway using a shared key. The strong swan instance within 371 AWS not only can connect to Azure but can also be used to facilitate 372 traffic to other nodes within the AWS VPC by configuring forwarding 373 and using appropriate routing rules for the VPC. Most Cloud 374 operators, such as AWS VPC or Azure VNET, use non-globally routable 375 CIDR from private IPv4 address ranges as specified by RFC1918. To 376 establish IPsec tunnel between two Cloud DCs, it is necessary to 377 exchange Public routable addresses for applications in different 378 Cloud DCs. [BGP-SDWAN] describes one method. Other methods are worth 379 exploring. 381 In summary, here are some approaches, available now (which might 382 change in the future), to interconnect workloads among different 383 Cloud DCs: 385 a) Utilize Cloud DC provided inter/intra-cloud connectivity 386 services (e.g., AWS Transit Gateway) to connect workloads 387 instantiated in multiple VPCs. Such services are provided with 388 the cloud gateway to connect to external networks (e.g., AWS 389 DirectConnect Gateway). 390 b) Hairpin all traffic through the customer gateway, meaning all 391 workloads are directly connected to the customer gateway, so 392 that communications among workloads within one Cloud DC must 393 traverse through the customer gateway. 394 c) Establish direct tunnels among different VPCs (AWS' Virtual 395 Private Clouds) and VNET (Azure's Virtual Networks) via 396 client's own virtual routers instantiated within Cloud DCs. 397 DMVPN (Dynamic Multipoint Virtual Private Network) or DSVPN 398 (Dynamic Smart VPN) techniques can be used to establish direct 399 Multi-point-to-Point or multi-point-to multi-point tunnels 400 among those client's own virtual routers. 402 Approach a) usually does not work if Cloud DCs are owned and managed 403 by different Cloud providers. 405 Approach b) creates additional transmission delay plus incurring 406 cost when exiting Cloud DCs. 408 For the Approach c), DMVPN or DSVPN use NHRP (Next Hop Resolution 409 Protocol) [RFC2735] so that spoke nodes can register their IP 410 addresses & WAN ports with the hub node. The IETF ION 411 (Internetworking over NBMA (non-broadcast multiple access) WG 412 standardized NHRP for connection-oriented NBMA network (such as ATM) 413 network address resolution more than two decades ago. 415 There are many differences between virtual routers in Public Cloud 416 DCs and the nodes in an NBMA network. NHRP cannot be used for 417 registering virtual routers in Cloud DCs unless an extension of such 418 protocols is developed for that purpose, e.g. taking NAT or dynamic 419 addresses into consideration. Therefore, DMVPN and/or DSVPN cannot 420 be used directly for connecting workloads in hybrid Cloud DCs. 422 Other protocols such as BGP can be used, as described in [BGP- 423 SDWAN]. 425 4.2. Desired Properties for Multi-Cloud Interconnection 427 Different Cloud Operators have different APIs to access their Cloud 428 resources. It is difficult to move applications built by one Cloud 429 operator's APIs to another. However, it is highly desirable to have 430 a single and consistent way to manage the networks and respective 431 security policies for interconnecting applications hosted in 432 different Cloud DCs. 434 The desired property would be having a single network fabric to 435 which different Cloud DCs and enterprise's multiple sites can be 436 attached or detached, with a common interface for setting desired 437 policies. SDWAN is positioned to become that network fabric enabling 438 Cloud DCs to be dynamically attached or detached. But the reality is 439 that different Cloud Operators have different access methods, and 440 Cloud DCs might be geographically far apart. More Cloud connectivity 441 problems are described in the subsequent sections. 443 The difficulty of connecting applications in different Clouds might 444 be stemmed from the fact that they are direct competitors. Usually 445 traffic flow out of Cloud DCs incur charges. Therefore, direct 446 communications between applications in different Cloud DCs can be 447 more expensive than intra Cloud communications. 449 5. Problems with MPLS-based VPNs extending to Hybrid Cloud DCs 451 Traditional MPLS-based VPNs have been widely deployed as an 452 effective way to support businesses and organizations that require 453 network performance and reliability. MPLS shifted the burden of 454 managing a VPN service from enterprises to service providers. The 455 CPEs attached to MPLS VPNs are also simpler and less expensive, 456 since they do not need to manage routes to remote sites; they simply 457 pass all outbound traffic to the MPLS VPN PEs to which the CPEs are 458 attached (albeit multi-homing scenarios require more processing 459 logic on CPEs). MPLS has addressed the problems of scale, 460 availability, and fast recovery from network faults, and 461 incorporated traffic-engineering capabilities. 463 However, traditional MPLS-based VPN solutions are sub-optimized for 464 connecting end-users to dynamic workloads/applications in cloud DCs 465 because: 467 - The Provider Edge (PE) nodes of the enterprise's VPNs might not 468 have direct connections to third party cloud DCs that are used 469 for hosting workloads with the goal of providing an easy access 470 to enterprises' end-users. 472 - It usually takes some time to deploy provider edge (PE) routers 473 at new locations. When enterprise's workloads are changed from 474 one cloud DC to another (i.e., removed from one DC and re- 475 instantiated to another location when demand changes), the 476 enterprise branch offices need to be connected to the new cloud 477 DC, but the network service provider might not have PEs located 478 at the new location. 480 One of the main drivers for moving workloads into the cloud is 481 the widely available cloud DCs at geographically diverse 482 locations, where apps can be instantiated so that they can be 483 as close to their end-users as possible. When the user base 484 changes, the applications may be migrated to a new cloud DC 485 location closest to the new user base. 487 - Most of the cloud DCs do not expose their internal networks. An 488 enterprise with a hybrid cloud deployment can use an MPLS-VPN 489 to connect to a Cloud provider at multiple locations. The 490 connection locations often correspond to gateways of different 491 Cloud DC locations from the Cloud provider. The different 492 Cloud DCs are interconnected by the Cloud provider's own 493 internal network. At each connection location (gateway), the 494 Cloud provider uses BGP to advertise all of the prefixes in the 495 enterprise's VPC, regardless of which Cloud DC a given prefix 496 is actually in. This can result in inefficient routing for the 497 end-to-end data path. 499 - Extensive usage of Overlay by Cloud DCs: 501 Many cloud DCs use an overlay to connect their gateways to the 502 workloads located inside the DC. There is currently no standard 503 that specifies the interworking between the Cloud Overlay and 504 the enterprise' existing underlay networks. One of the 505 characteristics of overlay networks is that some of the WAN 506 ports of the edge nodes connect to third party networks. There 507 is therefore a need to propagate WAN port information to remote 508 authorized peers in third party network domains in addition to 509 route propagation. Such an exchange cannot happen before 510 communication between peers is properly secured. 512 Another roadblock is the lack of a standard way to express and 513 enforce consistent security policies for workloads that not only use 514 virtual addresses, but in which are also very likely hosted in 515 different locations within the Cloud DC [RFC8192]. The current VPN 516 path computation and bandwidth allocation schemes may not be 517 flexible enough to address the need for enterprises to rapidly 518 connect to dynamically instantiated (or removed) workloads and 519 applications regardless of their location/nature (i.e., third party 520 cloud DCs). 522 6. Problem with using IPsec tunnels to Cloud DCs 523 As described in the previous section, many Cloud operators expose 524 their gateways for external entities (which can be enterprises 525 themselves) to directly establish IPsec tunnels. Enterprises can 526 also instantiate virtual routers within Cloud DCs to connect to 527 their on-premises devices via IPsec tunnels. If there is only one 528 enterprise location that needs to reach the Cloud DC, an IPsec 529 tunnel is a very convenient solution. 531 However, many medium-to-large enterprises usually have multiple 532 sites and multiple data centers. For workloads and apps hosted in 533 cloud DCs, multiple sites need to communicate securely with those 534 cloud workloads and apps. This section documents some of the issues 535 associated with using IPsec tunnels to connect enterprise premises 536 with cloud gateways. 538 6.1. Complexity of multi-point any-to-any interconnection 540 The dynamic workload instantiated in cloud DC needs to communicate 541 with multiple branch offices and on-premises data centers. Most 542 enterprises need multi-point interconnection among multiple 543 locations, which can be provided by means of MPLS L2/L3 VPNs. 545 Using IPsec overlay paths to connect all branches & on-premises data 546 centers to cloud DCs requires CPEs to manage routing among Cloud DCs 547 gateways and the CPEs located at other branch locations, which can 548 dramatically increase the complexity of the design, possibly at the 549 cost of jeopardizing the CPE performance. 551 The complexity of requiring CPEs to maintain routing among other 552 CPEs is one of the reasons why enterprises migrated from Frame Relay 553 based services to MPLS-based VPN services. 555 MPLS-based VPNs have their PEs directly connected to the CPEs. 556 Therefore, CPEs only need to forward all traffic to the directly 557 attached PEs, which are therefore responsible for enforcing the 558 routing policy within the corresponding VPNs. Even for multi-homed 559 CPEs, the CPEs only need to forward traffic among the directly 560 connected PEs. However, when using IPsec tunnels between CPEs and 561 Cloud DCs, the CPEs need to compute, select, establish and maintain 562 routes for traffic to be forwarded to Cloud DCs, to remote CPEs via 563 VPN, or directly. 565 6.2. Poor performance over long distance 567 When enterprise CPEs or gateways are far away from cloud DC gateways 568 or across country/continent boundaries, performance of IPsec tunnels 569 over the public Internet can be problematic and unpredictable. Even 570 though there are many monitoring tools available to measure delay 571 and various performance characteristics of the network, the 572 measurement for paths over the Internet is passive and past 573 measurements may not represent future performance. 575 Many cloud providers can replicate workloads in different available 576 zones. An App instantiated in a cloud DC closest to clients may have 577 to cooperate with another App (or its mirror image) in another 578 region or database server(s) in the on-premises DC. This kind of 579 coordination requires predicable networking behavior/performance 580 among those locations. 582 6.3. Scaling Issues with IPsec Tunnels 584 IPsec can achieve secure overlay connections between two locations 585 over any underlay network, e.g., between CPEs and Cloud DC Gateways. 587 If there is only one enterprise location connected to the cloud 588 gateway, a small number of IPsec tunnels can be configured on-demand 589 between the on-premises DC and the Cloud DC, which is an easy and 590 flexible solution. 592 However, for multiple enterprise locations to reach workloads hosted 593 in cloud DCs, the cloud DC gateway needs to maintain multiple IPsec 594 tunnels to all those locations (e.g., as a hub & spoke topology). 595 For a company with hundreds or thousands of locations, there could 596 be hundreds (or even thousands) of IPsec tunnels terminating at the 597 cloud DC gateway, which is not only very expensive (because Cloud 598 Operators usually charge their customers based on connections), but 599 can be very processing intensive for the gateway. Many cloud 600 operators only allow a limited number of (IPsec) tunnels & bandwidth 601 to each customer. Alternatively, you could use a solution like 602 group encryption where a single IPsec SA is necessary at the GW but 603 the drawback here is key distribution and maintenance of a key 604 server, etc. 606 7. Problems of Using SD-WAN to connect to Cloud DCs 607 SD-WAN can establish parallel paths over multiple underlay networks 608 between two locations on-demand, for example, to support the 609 connections established between two CPEs interconnected by a 610 traditional MPLS VPN ([RFC4364] or [RFC4664]) or by IPsec [RFC6071] 611 tunnels. 613 SD-WAN lets enterprises augment their current VPN network with cost- 614 effective, readily available Broadband Internet connectivity, 615 enabling some traffic offloading to paths over the Internet 616 according to differentiated, possibly application-based traffic 617 forwarding policies, or when the MPLS VPN connection between the two 618 locations is congested, or otherwise undesirable or unavailable. 620 7.1. SD-WAN among branch offices vs. interconnect to Cloud DCs 622 SD-WAN interconnection of branch offices is not as simple as it 623 appears. For an enterprise with multiple sites, using SD-WAN overlay 624 paths among sites requires each CPE to manage all the addresses that 625 local hosts have the potential to reach, i.e., map internal VPN 626 addresses to appropriate SD-WAN paths. This is similar to the 627 complexity of Frame Relay based VPNs, where each CPE needed to 628 maintain mesh routing for all destinations if they were to avoid an 629 extra hop through a hub router. Even though SD-WAN CPEs can get 630 assistance from a central controller (instead of running a routing 631 protocol) to resolve the mapping between destinations and SD-WAN 632 paths, SD-WAN CPEs are still responsible for routing table 633 maintenance as remote destinations change their attachments, e.g., 634 the dynamic workload in other DCs are de-commissioned or added. 636 Even though originally envisioned for interconnecting branch 637 offices, SD-WAN offers a very attractive way for enterprises to 638 connect to Cloud DCs. 640 The SD-WAN for interconnecting branch offices and the SD-WAN for 641 interconnecting to Cloud DCs have some differences: 643 - SD-WAN for interconnecting branch offices usually have two end- 644 points (e.g., CPEs) controlled by one entity (e.g., a 645 controller or management system operated by the enterprise). 646 - SD-WAN for Cloud DC interconnects may consider CPEs owned or 647 managed by the enterprise, while remote end-points are being 648 managed or controlled by Cloud DCs (For the ease of 649 description, let's call such CPEs asymmetrically-managed CPEs). 651 - Cloud DCs may have different entry points (or devices) with one 652 entry point that terminates a private direct connection (based 653 upon a leased line for example) and other entry points being 654 devices terminating the IPsec tunnels, as shown in Figure 2. 656 Therefore, the SD-WAN design becomes asymmetric. 657 +------------------------+ 658 | ,---. ,---. | 659 | (TN-1 ) ( TN-2)| TN: Tenant applications/workloads 660 | `-+-' +---+ `-+-' | 661 | +----|vR1|----+ | 662 | ++--+ | 663 | | +-+----+ 664 | | /Internet\ One path via 665 | +-------+ Gateway +---------------------+ 666 | \ / Internet \ 667 | +-+----+ \ 668 +------------------------+ \ 669 \ 670 +------------------------+ native traffic \ 671 | ,---. ,---. | without encryption| 672 | (TN-3 ) ( TN-4)| | 673 | `-+-' +--+ `-+-' | | +------+ 674 | +----|vR|-----+ | +----+ CPE | 675 | ++-+ | | +------+ 676 | | +-+----+ | 677 | | / virtual\ One path via IPsec Tunnel | 678 | +-------+ Gateway +-------------------------- + 679 | \ / Encrypted traffic over| 680 | +-+----+ public network | 681 +------------------------+ | 682 | 683 +------------------------+ | 684 | ,---. ,---. | Native traffic | 685 | (TN-5 ) ( TN-6)| without encryption | 686 | `-+-' +--+ `-+-' | over secure network| 687 | +----|vR|-----+ | | 688 | ++-+ | | 689 | | +-+----+ +------+ | 690 | | / \ Via Direct /customer\ | 691 | +-------+ Gateway +----------+ gateway |-----+ 692 | \ / Connect \ / 693 | +-+----+ +------+ 694 +------------------------+Customer GW has physical connection to AWS GW 696 Figure 2: Different Underlays to Reach Cloud DC 698 8. End-to-End Security Concerns for Data Flows 700 When IPsec tunnels established from enterprise on-premises CPEs 701 are terminated at the Cloud DC gateway where the workloads or 702 applications are hosted, some enterprises have concerns regarding 703 traffic to/from their workload being exposed to others behind the 704 data center gateway (e.g., exposed to other organizations that 705 have workloads in the same data center). 706 To ensure that traffic to/from workloads is not exposed to 707 unwanted entities, IPsec tunnels may go all the way to the 708 workload (servers, or VMs) within the DC. 710 9. Requirements for Dynamic Cloud Data Center VPNs 712 In order to address the aforementioned issues, any solution for 713 enterprise VPNs that includes connectivity to dynamic workloads or 714 applications in cloud data centers should satisfy a set of 715 requirements: 717 - The solution should allow enterprises to take advantage of the 718 current state-of-the-art in VPN technology, in both traditional 719 MPLS-based VPNs and IPsec-based VPNs (or any combination 720 thereof) that run over the public Internet. 721 - The solution should not require an enterprise to upgrade all 722 their existing CPEs. 723 - The solution should support scalable IPsec key management among 724 all nodes involved in DC interconnect schemes. 725 - The solution needs to support easy and fast, on-the-fly, VPN 726 connections to dynamic workloads and applications in third 727 party data centers, and easily allow these workloads to migrate 728 both within a data center and between data centers. 729 - Allow VPNs to provide bandwidth and other performance 730 guarantees. 731 - Be a cost-effective solution for enterprises to incorporate 732 dynamic cloud-based applications and workloads into their 733 existing VPN environment. 735 10. Security Considerations 737 The draft discusses security requirements as a part of the problem 738 space, particularly in sections 4, 5, and 8. 740 Solution drafts resulting from this work will address security 741 concerns inherent to the solution(s), including both protocol 742 aspects and the importance (for example) of securing workloads in 743 cloud DCs and the use of secure interconnection mechanisms. 745 11. IANA Considerations 747 This document requires no IANA actions. RFC Editor: Please remove 748 this section before publication. 750 12. References 752 12.1. Normative References 754 12.2. Informative References 756 [RFC2735] B. Fox, et al "NHRP Support for Virtual Private 757 networks". Dec. 1999. 759 [RFC8192] S. Hares, et al "Interface to Network Security Functions 760 (I2NSF) Problem Statement and Use Cases", July 2017 762 [ITU-T-X1036] ITU-T Recommendation X.1036, "Framework for creation, 763 storage, distribution and enforcement of policies for 764 network security", Nov 2007. 766 [RFC6071] S. Frankel and S. Krishnan, "IP Security (IPsec) and 767 Internet Key Exchange (IKE) Document Roadmap", Feb 2011. 769 [RFC4364] E. Rosen and Y. Rekhter, "BGP/MPLS IP Virtual Private 770 Networks (VPNs)", Feb 2006 772 [RFC4664] L. Andersson and E. Rosen, "Framework for Layer 2 Virtual 773 Private Networks (L2VPNs)", Sept 2006. 775 [BGP-SDWAN] L. Dunbar, et al. "BGP Extension for SDWAN Overlay 776 Networks", draft-dunbar-idr-bgp-sdwan-overlay-ext-03, 777 work-in-progress, Nov 2018. 779 13. Acknowledgments 781 Many thanks to Alia Atlas, Chris Bowers, Ignas Bagdonas, Michael 782 Huang, Liu Yuan Jiao, Katherine Zhao, and Jim Guichard for the 783 discussion and contributions. 785 Authors' Addresses 787 Linda Dunbar 788 Futurewei 789 Email: Linda.Dunbar@futurewei.com 791 Andrew G. Malis 792 Independent 793 Email: agmalis@gmail.com 795 Christian Jacquenet 796 Orange 797 Rennes, 35000 798 France 799 Email: Christian.jacquenet@orange.com 801 Mehmet Toy 802 Verizon 803 One Verizon Way 804 Basking Ridge, NJ 07920 805 Email: mehmet.toy@verizon.com