idnits 2.17.1 draft-dolson-sfc-hierarchical-03.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (October 2, 2015) is 3122 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Missing Reference: 'Virus' is mentioned on line 624, but not defined == Missing Reference: 'Monitor' is mentioned on line 636, but not defined == Missing Reference: 'CF' is mentioned on line 651, but not defined == Outdated reference: A later version (-05) exists of draft-homma-sfc-forwarding-methods-analysis-01 == Outdated reference: A later version (-11) exists of draft-ietf-sfc-architecture-07 == Outdated reference: A later version (-06) exists of draft-ietf-sfc-dc-use-cases-02 == Outdated reference: A later version (-28) exists of draft-ietf-sfc-nsh-00 Summary: 0 errors (**), 0 flaws (~~), 8 warnings (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Service Function Chaining D. Dolson 3 Internet-Draft Sandvine 4 Intended status: Informational S. Homma 5 Expires: April 4, 2016 NTT 6 D. Lopez 7 Telefonica I+D 8 M. Boucadair 9 Orange Group 10 D. Liu 11 Alibaba Group 12 October 2, 2015 14 Hierarchical Service Function Chaining 15 draft-dolson-sfc-hierarchical-03 17 Abstract 19 Hierarchical Service Function Chaining (hSFC) is a network 20 architecture allowing an organization to compartmentalize a large- 21 scale network into multiple domains of administration. 23 The goals of hSFC are to make a large-scale network easier to reason 24 about, simpler to control and to support independent functional 25 groups within large operators. 27 Status of This Memo 29 This Internet-Draft is submitted in full conformance with the 30 provisions of BCP 78 and BCP 79. 32 Internet-Drafts are working documents of the Internet Engineering 33 Task Force (IETF). Note that other groups may also distribute 34 working documents as Internet-Drafts. The list of current Internet- 35 Drafts is at http://datatracker.ietf.org/drafts/current/. 37 Internet-Drafts are draft documents valid for a maximum of six months 38 and may be updated, replaced, or obsoleted by other documents at any 39 time. It is inappropriate to use Internet-Drafts as reference 40 material or to cite them other than as "work in progress." 42 This Internet-Draft will expire on April 4, 2016. 44 Copyright Notice 46 Copyright (c) 2015 IETF Trust and the persons identified as the 47 document authors. All rights reserved. 49 This document is subject to BCP 78 and the IETF Trust's Legal 50 Provisions Relating to IETF Documents 51 (http://trustee.ietf.org/license-info) in effect on the date of 52 publication of this document. Please review these documents 53 carefully, as they describe your rights and restrictions with respect 54 to this document. Code Components extracted from this document must 55 include Simplified BSD License text as described in Section 4.e of 56 the Trust Legal Provisions and are provided without warranty as 57 described in the Simplified BSD License. 59 Table of Contents 61 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 62 1.1. Requirements Language . . . . . . . . . . . . . . . . . . 3 63 2. Hierarchical Service Function Chaining (hSFC) . . . . . . . . 4 64 2.1. Top Level . . . . . . . . . . . . . . . . . . . . . . . . 4 65 2.2. Lower Levels . . . . . . . . . . . . . . . . . . . . . . 5 66 3. Internal Boundary Node (IBN) . . . . . . . . . . . . . . . . 7 67 3.1. IBN Path Configuration . . . . . . . . . . . . . . . . . 7 68 3.1.1. Flow-Stateful IBN . . . . . . . . . . . . . . . . . . 7 69 3.1.2. Encoding Upper-Level Paths in Metadata . . . . . . . 9 70 3.1.3. Using Unique Paths per Upper-Level Path . . . . . . . 9 71 3.2. Gluing Levels Together . . . . . . . . . . . . . . . . . 10 72 4. Sub-domain Classifier . . . . . . . . . . . . . . . . . . . . 10 73 5. Control Plane Elements . . . . . . . . . . . . . . . . . . . 11 74 6. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 11 75 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 12 76 8. Security Considerations . . . . . . . . . . . . . . . . . . . 12 77 9. References . . . . . . . . . . . . . . . . . . . . . . . . . 12 78 9.1. Normative References . . . . . . . . . . . . . . . . . . 12 79 9.2. Informative References . . . . . . . . . . . . . . . . . 12 80 Appendix A. Examples of Hierarchical Service Function Chaining . 13 81 A.1. Reducing the Number of Service Function Paths . . . . . . 13 82 A.2. Managing a Distributed Data-Center Network . . . . . . . 15 83 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 17 85 1. Introduction 87 Service Function Chaining (SFC) is a technique for prescribing 88 differentiated traffic forwarding policies within the SFC domain. 89 SFC is described in detail in the SFC architecture document 90 [I-D.ietf-sfc-architecture], and is not repeated here. 92 In this document we consider the difficult problem of implementing 93 SFC across a large, geographically dispersed network comprised of 94 millions of hosts and thousands of network forwarding elements, 95 involving multiple operational teams (with varying functional 96 responsibilities). We expect asymmetrical routing is inherent in the 97 network, while recognizing that some Service Functions (SFs) require 98 bidirectional traffic for transport-layer sessions (e.g., NATs, 99 firewalls). We assume that some paths need to be selected on the 100 basis of application-specific data visible to the network, with 101 transport-layer coordinate (typically, 5-tuple) stickiness to 102 specific Service Function instances. 104 Note: in this document, the notion of the "path" of a packet is the 105 series of SF instances traversed by a packet. The means of 106 delivering packets between SFs (the forwarding mechanisms of the 107 underlay network) is not relevant to the current discussion. 109 Difficult problems are often made easier by decomposing them in a 110 hierarchical (nested) manner. So instead of considering an 111 omniscient SFC Control Plane that can manage (create, withdraw, 112 supervise, etc.) complete paths from one end of the network to the 113 other, we decompose the network into smaller sub-domains. Each sub- 114 domain may support a subset of the network applications or a subset 115 of the users. The criteria for determining decomposition into SFC- 116 enabled sub-domains are beyond the scope of this document. 118 Note that decomposing a network into multiple SFC-enabled domains 119 should permit end-to-end visibility of Service Functions and Service 120 Function Paths. Decomposition should also be implemented with 121 special care to ease monitoring and troubleshooting of the network as 122 a whole. 124 An example of simplifying a network by using multiple SF domains is 125 further discussed in [I-D.ietf-sfc-dc-use-cases]. 127 We assume the SF technology uses NSH [I-D.ietf-sfc-nsh] or a similar 128 labeling mechanism. 130 The "domains" discussed in this document are assumed to be under 131 control of a single organization, such that here is a strong trust 132 relationship between the domains. The intention of creating multiple 133 domains is to improve the ability to operate a network. It is 134 outside of the scope of the document to consider domains operated by 135 different organizations. 137 1.1. Requirements Language 139 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 140 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 141 document are to be interpreted as described in RFC 2119 [RFC2119]. 143 2. Hierarchical Service Function Chaining (hSFC) 145 A hierarchy has multiple levels. The top-most level encompasses the 146 entire network domain to be managed, and lower levels encompass 147 portions of the network. 149 2.1. Top Level 151 Considering the example in Figure 1, a top-level network domain 152 includes SFC components distributed over a wide area, including: 154 o classifiers (CFs), 156 o Service Function Forwarders (SFFs) and 158 o Sub-domains. 160 For the sake of clarity, components of the underlay network are not 161 shown; an underlay network is assumed to provide connectivity between 162 SFC components. 164 Top-level service function paths carry packets from classifiers 165 through a series of SFFs and sub-domains, with the operations within 166 sub-domains being opaque to the higher levels. 168 We expect the system to include a top-level control-plane having 169 responsibility for configuring forwarding and classification. The 170 top-level Service Chaining control-plane manages end-to-end service 171 chains and associated service function paths from network edge points 172 to sub-domains and configuring top-level classifiers at a coarse 173 level (e.g., based on source or destination host) to forward traffic 174 along paths that will transit appropriate sub-domains. The figure 175 shows one possible service chain passing from edge, through two sub- 176 domains, to network egress. The top-level control plane does NOT 177 configure classification or forwarding within the sub-domains. 179 At this network-wide level, the number of SFPs required is a linear 180 function of the number of ways in which a packet is required to 181 traverse different sub-domains and egress the network. Note that the 182 various paths which may be taken within a sub-domain are not 183 represented by distinct network-wide SFPs; specific policies at the 184 ingress nodes of each sub-domain bind flows to sub-domain paths. 186 Packets are classified at the edge of the network to select the paths 187 by which sub-domains are to be traversed. At the ingress of each 188 sub-domain, paths are reclassified to select the paths by which SFs 189 in the sub-domain are to be traversed. At the egress of each sub- 190 domain, packets are returned to the top-level paths. Contrast this 191 with an approach requiring the top-level classifier to select paths 192 to specify all of the SFs in each sub-domain. 194 It should be assumed that some service functions in the network 195 require bidirectional symmetry of paths (see more in Section 4). 196 Therefore the classifiers at the top level must be configured with 197 policies ensuring server-to-client packets take the reverse path of 198 client-to-server packet through sub-domains. (Recall the "path" 199 denotes the series of service functions; the precise physical network 200 path within the underlay network is not relevant here.) 202 +------------+ 203 |Sub-domain#1| 204 | in DC1 | 205 +----+-------+ 206 | 207 .---- SFF1 ------. +--+ 208 +--+ / / | \--|CF| 209 --->|CF|--/---->' | \ +--+ 210 +--+ / SC#1 | \ 211 | | | 212 | V .------>|---> 213 | / / | 214 \ | / / 215 +--+ \ | / / +--+ 216 |CF|---\ | / /---|CF| 217 +--+ '---- SFF2 ------' +--+ 218 | 219 +----+-------+ 220 |Sub-domain#2| 221 | in DC2 | 222 +------------+ 224 One path is shown from edge classifier to SFF1 to Sub-domain#1 225 (residing in data-center1) to SFF1 to SFF2 (residing in data-center 226 2) to Sub-domain#2 to SFF2 to network egress. 228 Figure 1: Network-wide view of Top Level of Hierarchy 230 2.2. Lower Levels 232 Each of the sub-domains in Figure 1 is an SFC domain. 234 Unlike the top level, however, data packets entering the sub-domain 235 are already encapsulated within SFC transport. Figure 2 shows a sub- 236 domain interfaced with a higher-level domain by means of an Internal 237 Boundary Node (IBN). It is the purpose of the IBN to remove packets 238 from the SFC encapsulation, apply Classification rules, and direct 239 the packets to the selected local service function paths terminating 240 at an egress IBN. The egress SFC Domain Gateway finally restores 241 packets to the original SFC transport and hands them off to SFFs. 243 Each sub-domain intersects a subset of the total paths that are 244 possible in the higher-level domain. An IBN is concerned with 245 higher-level paths, but only those traversing the sub-domain. A top- 246 level controller may configure the IBN as an SF (the IBN plays the SF 247 role in the top-level domain). 249 We expect each sub-domain to have a control-plane that can operate 250 independently of the top-level control-plane. The sub-domain 251 control-plane configures the classification and forwarding rules in 252 the sub-domain. The classification rules reside in the IBN, where 253 packets are moved from SFC encapsulation of the top-level domain to 254 and from SFC encapsulation of the lower-level domain. 256 +----+ +-----+ +----------------------+ +-----+ 257 | |SC#1| SFF | | IBN 1 | | SFF | 258 ->| |================* *===============> 259 | | +-----+ | # (in DC 1) # | +-----+ 260 | CF | | V # | 261 | | |+---+ +---+| Top domain 262 | | * * * * *||CF | * * * * * *|SFF|| * * * * * 263 | | * |+---+ +-+-+| * 264 +----+ * | | | | | | Sub * 265 * +-o-o--------------o-o-+ domain* 266 * SC#2 | |SC#1 ^ ^ #1 * 267 * +-----+ | | | * 268 * | V | | * 269 * | +---+ +------+ | | * 270 * | |SFF|->|SF#1.1|--+ | * 271 * | +---+ +------+ | * 272 * V | * 273 * +---+ +------+ +---+ +------+ * 274 * |SFF|->|SF#2.1|->|SFF|->|SF#2.2| * 275 * +---+ +------+ +---+ +------+ * 276 * * * * * * * * * * * * * * * * * * * * * * 278 *** Sub-domain boundary; === top-level chain; --- low-level chain. 280 Figure 2: Sub-domain within a higher-level domain 282 If desired, the pattern can be applied recursively. For example, 283 SF#1.1 in Figure 2 could be a sub-domain of the sub-domain. 285 3. Internal Boundary Node (IBN) 287 A network element termed "Internal Boundary Node" (IBN) bridges 288 packets between domains. It looks like an SF to the higher level, 289 and looks like a classifier and end-of-chain to the lower level. 291 To achieve the benefits of hierarchy, the IBN should be applying more 292 granular traffic classification rules at the lower level than the 293 traffic passed to it. This means that the number of SF Paths within 294 the lower level is greater than the number of SF Paths arriving to 295 the IBN. 297 The IBN is also the termination of lower-level SF paths. This is 298 because the packets exiting lower-level SF paths must be returned to 299 the higher-level SF paths and forwarded to the next hop in the 300 higher-level domain. 302 3.1. IBN Path Configuration 304 An operator of a lower-level SF Domain may be aware of which high- 305 level paths transit their domain, or they may wish to accept any 306 paths. 308 When packets enter the sub-domain, the Path Identifier and Path Index 309 are re-marked according to the path selected by the classifier. 311 After exiting a path in the sub-domain, packets can be restored to an 312 upper-level SF path by these methods: 314 1. Stateful per flow, 316 2. Pushing path identifier into metadata, 318 3. Using unique lower-level paths per upper-level path. 320 3.1.1. Flow-Stateful IBN 322 An IBN can be flow-aware, returning packets to the correct higher- 323 level SF path on the basis of the transport-layer coordinates 324 (typically, a 5-tuple) of packets exiting the lower-level SF paths. 326 When packets are received by the IBN on a higher-level path, the 327 encapsulated packets are parsed for IP and transport-layer (TCP, 328 UDP...) coordinates. State is created, indexed by these coordinates 329 (a 5-tuple of {source-IP, destination-IP, source-port, destination- 330 port and transport protocol} in a typical case). The state contains 331 critical fields of the encapsulating SFC header (or perhaps the 332 entire header). 334 The simplest approach has the packets return to the same IBN at the 335 end of the chain that classified the packet at the start of the 336 chain. This is because the required transport-coordinates state is 337 rapidly changing and most efficiently kept locally. If the packet is 338 returned to a different IBN for egress, transport-coordinates state 339 must be synchronized between the IBNs. 341 When a packet returns to the IBN at the end of a chain, the SFC 342 header is removed, the packet is parsed for IP and transport-layer 343 coordinates, and state is retrieved from them. The state contains 344 the information required to forward the packet within the higher- 345 level service chain. 347 State cannot be created by packets arriving from the lower-level 348 chain; when state cannot be found for such packets, they MUST be 349 dropped. 351 This stateful approach is limited to use with SFs that retain the 352 transport coordinates of the packet. This approach cannot be used 353 with SFs that modify those coordinates (e.g., as done by a NAT) or 354 otherwise create packets for new coordinates other than those 355 received (e.g., as an HTTP cache might do to retrieve content on 356 behalf of the original flow). In both cases, the fundamental problem 357 is the inability to forward packets when state cannot be found for 358 the packet transport-layer coordinates. 360 In the stateful approach, there are issues caused by the state, such 361 as how long the state should be maintained (it MUST time out 362 eventually), as well as whether the state needs to be replicated to 363 other devices to create a highly available network. 365 It is valid to consider the state disposable after failure, since it 366 can be re-created by each new packet arriving from the higher-level 367 domain. For example, if an IBN loses all flow state, the state is 368 re-created by an end-point retransmitting a TCP packet. 370 If an SFC domain handles multiple network regions (e.g., multiple 371 private networks), the coordinates may be augmented with additional 372 parameters, perhaps using some metadata to identify the network 373 region. 375 In this stateful approach, it is not necessary for the sub-domain's 376 control-plane to modify paths when higher-level paths are changed. 377 The complexity of the higher-level domain does not cause complexity 378 in the lower-level domain. 380 3.1.2. Encoding Upper-Level Paths in Metadata 382 An IBN can push the upper-level service path identifier (SPI) and 383 service index (SI) (or encoding thereof) into a metadata field of the 384 lower-level encapsulation (e.g., placing upper-level path information 385 into a metadata field of NSH). When packets exit the lower-level 386 path, the upper-level SPI and SI can be restored from the metadata 387 retrieved from the packet. 389 This approach requires the SFs in the path to be capable of 390 forwarding the metadata and appropriately attaching metadata to any 391 packets injected for a flow. 393 Using new metadata may inflate packet size when variable-length 394 metadata (type 2 from NSH [I-D.ietf-sfc-nsh]) is used. 396 It is conceivable that the MD-type 1 Mandatory Context Header fields 397 of NSH [I-D.ietf-sfc-nsh] are not all relevant to the lower-level 398 domain. In this case, one of the metadata slots of the Mandatory 399 Context Header could be repurposed within the lower-level domain, and 400 restored when leaving. 402 In this metadata approach, it is not necessary for the sub-domain's 403 controller to modify paths when higher-level paths are changed. The 404 complexity of the higher-level domain does not cause complexity in 405 the lower-level domain. 407 3.1.3. Using Unique Paths per Upper-Level Path 409 In this approach, paths within the sub-domain are constrained so that 410 a path identifier (of the sub-domain) unambiguously indicates the 411 egress path (of the upper domain). This allows the original path 412 information to be restored at sub-domain egress from a look-up table 413 using the sub-domain path identifier. 415 Whenever the upper-level domain provisions a path via the lower-level 416 domain, the lower-level domain controller must provision 417 corresponding paths to traverse the lower-level domain. 419 A down-side of this approach is that the number of paths in the 420 lower-level domain is multiplied by the number of paths in the 421 higher-level domain that traverse the lower-level domain. I.e., a 422 sub-path must be created for each combination of upper Path 423 identifier and lower path. 425 3.2. Gluing Levels Together 427 The path identifier or metadata on a packet received by the IBN may 428 be used as input to reclassification and path selection within the 429 lower-level domain. 431 In some cases the meanings of the various path IDs and metadata must 432 be coordinated between domains. 434 One approach is to use well-known identifier values in metadata, 435 communicated by some organizational registry. 437 Another approach is to use well-known labels for path identifiers or 438 metadata, as an indirection to the actual identifiers. The actual 439 identifiers can be assigned by control-plane systems. For example, a 440 sub-domain classifier could have a policy, "if pathID=classA then 441 chain packet to path 1234"; the higher-level controller would be 442 expected to configure the concrete higher-level pathID for classA. 444 4. Sub-domain Classifier 446 Within the sub-domain (referring to Figure 2), after the IBN removes 447 higher-level encapsulation from incoming packets, it sends the 448 packets to the classifier, which selects the encapsulation for the 449 packet within the sub-domain. 451 One of the goals of the hierarchical approach is to make it easy to 452 have transport-flow-aware service chaining with bidirectional paths. 453 For example, it is desired that for each TCP flow, the client-to- 454 server packets traverse the same SFs as the server-to-client packets, 455 but in the opposite sequence. We call this bidirectional symmetry. 456 If bidirectional symmetry is required, it is the responsibility of 457 the control-plane to be aware of symmetric paths and configure the 458 classifier to chain the traffic in a symmetric manner. 460 Another goal of the hierarchical approach is to simplify the 461 mechanisms of scaling in and scaling out service functions. All of 462 the complexities of load-balancing among multiple SFs can be handled 463 within a sub-domain, under control of the classifier, allowing the 464 higher-level domain to be oblivious to the existence of multiple SF 465 instances. 467 Considering the requirements of bidirectional symmetry and load- 468 balancing, it is useful to have all packets entering a sub-domain to 469 be received by the same classifier or a coordinated cluster of 470 classifiers. There are both stateful and stateless approaches to 471 ensuring bidirectional symmetry. 473 5. Control Plane Elements 475 Controllers have been mentioned in this document without much 476 explanation. Although control protocols have not yet been 477 standardized, from the point of view of hierarchical service function 478 chaining we have these expectations: 480 o Each control-plane instance manages a single level of hierarchy of 481 a single domain. 483 o Each control-plane is agnostic about other levels of hierarchy. 484 This aspect allows humans to reason about the system within a 485 single domain and allows control-plane algorithms to use only 486 domain-local inputs. Top-level control does not need visibility 487 to sub-domain policies, nor does sub-domain control need 488 visibility to higher-level policies. 490 o Sub-domain control-planes are agnostic about control-planes of 491 other sub-domains. This allows both humans and machines to 492 manipulate sub-domain policy without considering policies of other 493 domains. 495 Recall that the IBN acts as an SF in the higher-level domain 496 (receiving SF instructions from the higher-level control-plane) and 497 as a classifier in the lower-level domain (receiving classification 498 rules from the sub-domain control-plane). In this view, it is the 499 IBN that glues the layers together. 501 The above expectations are not intended to prohibit network-wide 502 control. A control hierarchy can be envisaged to distribute 503 information and instructions to multiple domains and sub-domains. 504 Control hierarchy is outside the scope of this document. 506 6. Acknowledgements 508 The concept of Hierarchical Service Path Domains was introduced in 509 draft-homma-sfc-forwarding-methods-analysis-01 510 [I-D.homma-sfc-forwarding-methods-analysis] as a means to improve 511 scalability of service chaining in large networks. 513 The authors would like to thank the following individuals for taking 514 the time to read and provide valuable feedback: 516 Ron Parker 518 Christian Jacquenet 520 Jie Cao 522 7. IANA Considerations 524 This memo includes no request to IANA. 526 8. Security Considerations 528 Hierarchical service function chaining makes use of service chaining 529 architecture, and hence inherits the security considerations 530 described in the architecture document. 532 Furthermore, hierarchical service function chaining inherits security 533 considerations of the data-plane protocols (e.g., NSH) and control- 534 plane protocols used to realize the solution. 536 The systems described in this document bear responsibility for 537 forwarding internet traffic. In some cases the systems are 538 responsible for maintaining separation of traffic in private 539 networks. 541 This document describes systems within different domains of 542 administration that must have consistent configurations in order to 543 properly forward traffic and to maintain private network separation. 544 Any protocol designed to distribute the configurations must be secure 545 from tampering. 547 All of the systems and protocols must be secure from modification by 548 untrusted agents. 550 9. References 552 9.1. Normative References 554 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 555 Requirement Levels", BCP 14, RFC 2119, 556 DOI 10.17487/RFC2119, March 1997, 557 . 559 9.2. Informative References 561 [I-D.homma-sfc-forwarding-methods-analysis] 562 Homma, S., Kengo, K., Lopez, D., Stiemerling, M., and D. 563 Dolson, "Analysis on Forwarding Methods for Service 564 Chaining", draft-homma-sfc-forwarding-methods-analysis-01 565 (work in progress), January 2015. 567 [I-D.ietf-sfc-architecture] 568 Halpern, J. and C. Pignataro, "Service Function Chaining 569 (SFC) Architecture", draft-ietf-sfc-architecture-07 (work 570 in progress), March 2015. 572 [I-D.ietf-sfc-dc-use-cases] 573 Surendra, S., Tufail, M., Majee, S., Captari, C., and S. 574 Homma, "Service Function Chaining Use Cases In Data 575 Centers", draft-ietf-sfc-dc-use-cases-02 (work in 576 progress), January 2015. 578 [I-D.ietf-sfc-nsh] 579 Quinn, P. and U. Elzur, "Network Service Header", draft- 580 ietf-sfc-nsh-00 (work in progress), March 2015. 582 Appendix A. Examples of Hierarchical Service Function Chaining 584 The advantage of hierarchical service function chaining compared with 585 normal or flat service function chaining is that it can reduce the 586 management complexity significantly. This section discusses examples 587 that show the advantage of hierarchical service function chaining. 589 A.1. Reducing the Number of Service Function Paths 591 In this case, hierarchical service function chaining is used to 592 simplify service function chaining management by reducing the number 593 of Service Function Paths. 595 As shown in Figure 3, there are two domains each with different 596 concerns: a Security Domain that selects Service Functions based on 597 network conditions and an Optimization Domain that selects Service 598 Functions based on traffic protocol. 600 There are five security functions deployed in the Security Domain. 601 The Security Domain operator wants to enforce the five different 602 security policies, and the Optimization Domain operator wants to 603 apply different optimizations (either cache or video optimization) to 604 each of these two types of traffic. If we use flat SFC (normal 605 branching), 10 SFPs are needed in each domain. In contrast, if we 606 use hierarchical SFC, only 5 SFPs in Security Domain and 2 SFPs in 607 Optimization Domain will be required, as shown in Figure 4. 609 In the flat model, the number of SFPs is the product of the number of 610 functions in all of the domains. In the hSFC model, the number of 611 SFPs is the sum of the number of functions. For example, adding a 612 "bypass" path in the Optimization Domain would cause the flat model 613 to require 15 paths (5 more), but cause the hSFC model to require one 614 more path in the Optimization Domain. 616 . . . . . . . . . . . . . . . . . . . . . . . . . 617 . Security Domain . . Optimization Domain . 618 . . . . 619 . +-1---[ ]----------------->[Cache ]-------> 620 . | [ WAF ] . . . 621 . +-2-->[ ]----------------->[Video Opt.]----> 622 . | . . . 623 . +-3---[Anti ]----------------->[Cache ]-------> 624 . | [Virus] . . . 625 . +-4-->[ ]----------------->[Video Opt.]----> 626 . | . . . 627 . +-5-->[ ]----------------->[Cache ]-------> 628 [DPI]--->[CF]---| [ IPS ] . . . 629 . +-6-->[ ]----------------->[Video Opt.]----> 630 . | . . . 631 . +-7-->[ ]----------------->[Cache ]-------> 632 . | [ IDS ] . . . 633 . +-8-->[ ]----------------->[Video Opt.]----> 634 . | . . . 635 . +-9-->[Traffic]--------------->[Cache ]-------> 636 . | [Monitor] . . . 637 . +-10->[ ]--------------->[Video Opt.]----> 638 . . . . . . . . . . . . . . . . . . . . . . . . . 640 The classifier must select paths that determine the combination of 641 Security and Optimization concerns. 1:WAF+Cache, 2:WAF+VideoOpt, 642 3:AntiVirus+Cache, 4:AntiVirus+VideoOpt, 5: IPS+Cache, 643 6:IPS+VideoOpt, 7:IDS+Cache, 8:IDS+VideoOpt, 9:TrafficMonitor+Cache, 644 10:TrafficMonitor+VideoOpt 646 Figure 3: Flat SFC (Normal Branching) 648 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 649 . Security Domain . . Optimization Domain . 650 . . . . 651 [CF]---->[ [CF] IBN ]---------->[ [CF] IBN ]----> 652 . | ^ . . | ^ . 653 . +----->[ WAF ]-----+ . . +-->[ Cache ]---------+ . 654 . | | . . | | . 655 . +-->[Anti-Virus]---+ . . +-->[Video Opt]-------+ . 656 . | | . . . 657 . +----->[ IPS ]-----+ . . . . . . . . . . . . . . . . 658 . | | . 659 . +----->[ IDS ]-----+ . 660 . | | . 661 . +-->[ Traffic ]----+ . 662 . [ Monitor ] . 663 . . . . . . . . . . . . . . . 665 Figure 4: Simplified Path Management with Hierarchical SFC 667 A.2. Managing a Distributed Data-Center Network 669 Hierarchical service function chaining can be used to simplify inter- 670 data-center SFC management. In the example of Figure 5, shown below, 671 there is a central data center (Central DC) and multiple local data 672 centers (Local DC#1, #2, #3) that are deployed in a geographically 673 distributed manner. All of the data centers are under a single 674 administrative domain. 676 The central DC may have some service functions that the local DC 677 needs, such that the local DC needs to chain traffic via the central 678 DC. This could be because: 680 o Some service functions are deployed as dedicated hardware 681 appliances, and there is a desire to lower the cost (both CAPEX 682 and OPEX) of deploying such service functions in all data centers. 684 o Consider the case when service functions are being trialed, 685 introduced or otherwise handle a relatively small amount of 686 traffic. It may be cheaper to manage these service functions in a 687 single central data center and steer packets to the central data 688 center than to manage these service functions in all data centers. 690 +-----------+ 691 |Central DC | 692 +-----------+ 693 ^ ^ ^ 694 | | | 695 .---|--|---|----. 696 / / | | \ 697 / / | \ \ 698 +-----+ / / | \ \ +-----+ 699 |Local| | / | \ | |Local| 700 |DC#1 |--|--. | .----|----|DC#3 | 701 +-----+ | | | +-----+ 702 \ | / 703 \ | / 704 \ | / 705 '----------------' 706 | 707 +-----+ 708 |Local| 709 |DC#2 | 710 +-----+ 712 Figure 5: Simplify Inter-DC SFC Management 714 For large data center operators, one local DC may have tens of 715 thousands of servers and hundred of thousands of virtual machines. 716 SFC can be used to manage user traffic. For example, SFC can be used 717 to classify user traffic based on service type, DDoS state etc. 719 In such large scale data center, using flat SFC is very complex, 720 requiring a super-controller to configure all data centers. For 721 example, any changes to Service Functions or Service Function Paths 722 in the central DC (e.g., deploying a new SF) would require updates to 723 all of the Service Function Paths in the local DCs accordingly. 724 Furthermore, requirements for symmetric paths add additional 725 complexity when flat SFC is used in this scenario. 727 Conversely, if using hierarchical SFC, each data center can be 728 managed independently and the management complexity could be reduced 729 significantly. Service Function Paths between data centers can 730 represent abstract notions without regard to details within data 731 centers. Independent controllers can be used for the top level 732 (getting packets to pass the correct data centers) and local levels 733 (getting packets to specific SF instances). 735 Authors' Addresses 737 David Dolson 738 Sandvine 739 408 Albert Street 740 Waterloo, ON N2L 3V3 741 Canada 743 Phone: +1 519 880 2400 744 Email: ddolson@sandvine.com 746 Shunsuke Homma 747 NTT, Corp. 748 3-9-11, Midori-cho 749 Musashino-shi, Tokyo 180-8585 750 Japan 752 Email: homma.shunsuke@lab.ntt.co.jp 754 Diego R. Lopez 755 Telefonica I+D 756 Don Ramon de la Cruz, 82 757 Madrid 28006 758 Spain 760 Phone: +34 913 129 041 761 Email: diego.r.lopez@telefonica.com 763 Mohamed Boucadair 764 Orange Group 765 Rennes 35000 766 France 768 Email: mohamed.boucadair@orange.com 770 Dapeng Liu 771 Alibaba Group 772 Beijing 100022 773 China 775 Email: max.ldp@alibaba-inc.com