idnits 2.17.1 draft-ietf-anima-autonomic-control-plane-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 : ---------------------------------------------------------------------------- ** The document seems to lack separate sections for Informative/Normative References. All references will be assumed normative when checking for downward references. ** The document seems to lack a both a reference to RFC 2119 and the recommended RFC 2119 boilerplate, even if it appears to use RFC 2119 keywords. RFC 2119 keyword, line 251: '... 1. The ACP SHOULD provide robust c...' RFC 2119 keyword, line 256: '... 2. The ACP MUST have a separate ad...' RFC 2119 keyword, line 260: '... 3. The ACP MUST use autonomically ...' RFC 2119 keyword, line 265: '... 4. The ACP MUST be generic. Usabl...' RFC 2119 keyword, line 266: '...rastructure. It MUST NOT be tied to a...' (19 more instances...) Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (March 21, 2016) is 2958 days in the past. Is this intentional? Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) == Missing Reference: 'TBD' is mentioned on line 883, but not defined == Unused Reference: 'I-D.behringer-anima-autonomic-addressing' is defined on line 1303, but no explicit reference was found in the text == Outdated reference: A later version (-03) exists of draft-behringer-anima-autonomic-addressing-02 ** Downref: Normative reference to an Informational draft: draft-behringer-anima-reference-model (ref. 'I-D.behringer-anima-reference-model') ** Downref: Normative reference to an Informational draft: draft-behringer-autonomic-control-plane (ref. 'I-D.behringer-autonomic-control-plane') ** Downref: Normative reference to an Informational draft: draft-eckert-anima-stable-connectivity (ref. 'I-D.eckert-anima-stable-connectivity') == Outdated reference: A later version (-45) exists of draft-ietf-anima-bootstrapping-keyinfra-02 == Outdated reference: A later version (-15) exists of draft-ietf-anima-grasp-04 ** Obsolete normative reference: RFC 6347 (Obsoleted by RFC 9147) ** Downref: Normative reference to an Informational RFC: RFC 7404 ** Downref: Normative reference to an Informational RFC: RFC 7575 ** Downref: Normative reference to an Informational RFC: RFC 7576 Summary: 9 errors (**), 0 flaws (~~), 6 warnings (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 ANIMA WG M. Behringer, Ed. 3 Internet-Draft S. Bjarnason 4 Intended status: Standards Track Balaji. BL 5 Expires: September 22, 2016 T. Eckert 6 Cisco Systems 7 March 21, 2016 9 An Autonomic Control Plane 10 draft-ietf-anima-autonomic-control-plane-02 12 Abstract 14 Autonomic functions need a control plane to communicate, which 15 depends on some addressing and routing. This Autonomic Control Plane 16 should ideally be self-managing, and as independent as possible of 17 configuration. This document defines an "Autonomic Control Plane", 18 with the primary use as a control plane for autonomic functions. It 19 also serves as a "virtual out of band channel" for OAM communications 20 over a network that is not configured, or mis-configured. 22 Status of This Memo 24 This Internet-Draft is submitted in full conformance with the 25 provisions of BCP 78 and BCP 79. 27 Internet-Drafts are working documents of the Internet Engineering 28 Task Force (IETF). Note that other groups may also distribute 29 working documents as Internet-Drafts. The list of current Internet- 30 Drafts is at http://datatracker.ietf.org/drafts/current/. 32 Internet-Drafts are draft documents valid for a maximum of six months 33 and may be updated, replaced, or obsoleted by other documents at any 34 time. It is inappropriate to use Internet-Drafts as reference 35 material or to cite them other than as "work in progress." 37 This Internet-Draft will expire on September 22, 2016. 39 Copyright Notice 41 Copyright (c) 2016 IETF Trust and the persons identified as the 42 document authors. All rights reserved. 44 This document is subject to BCP 78 and the IETF Trust's Legal 45 Provisions Relating to IETF Documents 46 (http://trustee.ietf.org/license-info) in effect on the date of 47 publication of this document. Please review these documents 48 carefully, as they describe your rights and restrictions with respect 49 to this document. Code Components extracted from this document must 50 include Simplified BSD License text as described in Section 4.e of 51 the Trust Legal Provisions and are provided without warranty as 52 described in the Simplified BSD License. 54 Table of Contents 56 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 57 2. Use Cases for an Autonomic Control Plane . . . . . . . . . . 4 58 2.1. An Infrastructure for Autonomic Functions . . . . . . . . 4 59 2.2. Secure Bootstrap over an Unconfigured Network . . . . . . 4 60 2.3. Data Plane Independent Permanent Reachability . . . . . . 5 61 3. Requirements . . . . . . . . . . . . . . . . . . . . . . . . 6 62 4. Overview . . . . . . . . . . . . . . . . . . . . . . . . . . 6 63 5. Self-Creation of an Autonomic Control Plane . . . . . . . . . 8 64 5.1. Preconditions . . . . . . . . . . . . . . . . . . . . . . 8 65 5.2. Candidate ACP Neighbor Selection . . . . . . . . . . . . 8 66 5.3. Capability Negotiation . . . . . . . . . . . . . . . . . 9 67 5.4. Channel Establishment . . . . . . . . . . . . . . . . . . 10 68 5.5. Context Separation . . . . . . . . . . . . . . . . . . . 10 69 5.6. Addressing inside the ACP . . . . . . . . . . . . . . . . 11 70 5.6.1. Fundamental Concepts of Autonomic Addressing . . . . 11 71 5.6.2. The Base Addressing Scheme . . . . . . . . . . . . . 12 72 5.6.3. Possible Sub-Schemes . . . . . . . . . . . . . . . . 13 73 5.6.4. Usage of the Zone Field . . . . . . . . . . . . . . . 14 74 5.7. Routing in the ACP . . . . . . . . . . . . . . . . . . . 15 75 6. Workarounds for Non-Autonomic Nodes . . . . . . . . . . . . . 16 76 6.1. Connecting a Non-Autonomic Controller / NMS system . . . 16 77 6.2. ACP through Non-Autonomic L3 Clouds . . . . . . . . . . . 16 78 7. Building the ACP . . . . . . . . . . . . . . . . . . . . . . 17 79 7.1. Neighbor discovery via GRASP . . . . . . . . . . . . . . 17 80 7.2. Channel Selection . . . . . . . . . . . . . . . . . . . . 17 81 7.3. Security Association protocols . . . . . . . . . . . . . 18 82 7.3.1. ACP via IPsec . . . . . . . . . . . . . . . . . . . . 18 83 7.3.2. ACP via GRE/IPsec . . . . . . . . . . . . . . . . . . 19 84 7.3.3. ACP via dTLS . . . . . . . . . . . . . . . . . . . . 19 85 7.3.4. GRASP/TLS negotiation . . . . . . . . . . . . . . . . 19 86 7.3.5. ACP Security Profiles . . . . . . . . . . . . . . . . 20 87 7.4. GRASP instance details . . . . . . . . . . . . . . . . . 20 88 8. Self-Healing Properties . . . . . . . . . . . . . . . . . . . 21 89 9. Self-Protection Properties . . . . . . . . . . . . . . . . . 22 90 10. The Administrator View . . . . . . . . . . . . . . . . . . . 22 91 11. Explanations . . . . . . . . . . . . . . . . . . . . . . . . 23 92 11.1. Why GRASP to discover autonomic neighbors . . . . . . . 23 93 12. Security Considerations . . . . . . . . . . . . . . . . . . . 24 94 13. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 25 95 14. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 25 96 15. Change log [RFC Editor: Please remove] . . . . . . . . . . . 26 97 15.1. Initial version . . . . . . . . . . . . . . . . . . . . 26 98 15.2. draft-behringer-anima-autonomic-control-plane-00 . . . . 26 99 15.3. draft-behringer-anima-autonomic-control-plane-01 . . . . 26 100 15.4. draft-behringer-anima-autonomic-control-plane-02 . . . . 26 101 15.5. draft-behringer-anima-autonomic-control-plane-03 . . . . 27 102 15.6. draft-ietf-anima-autonomic-control-plane-00 . . . . . . 27 103 15.7. draft-ietf-anima-autonomic-control-plane-01 . . . . . . 27 104 15.8. draft-ietf-anima-autonomic-control-plane-02 . . . . . . 28 105 16. References . . . . . . . . . . . . . . . . . . . . . . . . . 28 106 Appendix A. Background on the choice of routing protocol . . . . 29 107 Appendix B. Alternative: An ACP without Separation . . . . . . . 31 108 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 31 110 1. Introduction 112 Autonomic Networking is a concept of self-management: Autonomic 113 functions self-configure, and negotiate parameters and settings 114 across the network. [RFC7575] defines the fundamental ideas and 115 design goals of Autonomic Networking. A gap analysis of Autonomic 116 Networking is given in [RFC7576]. The reference architecture for 117 Autonomic Networking in the IETF is currently being defined in the 118 document [I-D.behringer-anima-reference-model] 120 Autonomic functions need a stable and robust infrastructure to 121 communicate on. This infrastructure should be as robust as possible, 122 and it should be re-usable by all autonomic functions. [RFC7575] 123 calls it the "Autonomic Control Plane". This document defines the 124 Autonomic Control Plane. 126 Today, the management and control plane of networks typically runs in 127 the global routing table, which is dependent on correct configuration 128 and routing. Misconfigurations or routing problems can therefore 129 disrupt management and control channels. Traditionally, an out of 130 band network has been used to recover from such problems, or 131 personnel is sent on site to access devices through console ports. 132 However, both options are operationally expensive. 134 In increasingly automated networks either controllers or distributed 135 autonomic service agents in the network require a control plane which 136 is independent of the network they manage, to avoid impacting their 137 own operations. 139 This document describes options for a self-forming, self-managing and 140 self-protecting "Autonomic Control Plane" (ACP) which is inband on 141 the network, yet as independent as possible of configuration, 142 addressing and routing problems (for details how this achieved, see 143 Section 5). It therefore remains operational even in the presence of 144 configuration errors, addressing or routing issues, or where policy 145 could inadvertently affect control plane connectivity. The Autonomic 146 Control Plane serves several purposes at the same time: 148 o Autonomic functions communicate over the ACP. The ACP therefore 149 supports directly Autonomic Networking functions, as described in 150 [I-D.behringer-anima-reference-model]. For example, GRASP 151 [I-D.ietf-anima-grasp] can run inside the ACP. 153 o An operator can use it to log into remote devices, even if the 154 data plane is misconfigured or unconfigured. 156 o A controller or network management system can use it to securely 157 bootstrap network devices in remote locations, even if the network 158 in between is not yet configured; no data-plane dependent 159 bootstrap configuration is required. An example of such a secure 160 bootstrap process is described in 161 [I-D.ietf-anima-bootstrapping-keyinfra] 163 This document describes some use cases for the ACP in Section 2, it 164 defines the requirements in Section 3, Section 4 gives an overview 165 how an Autonomic Control Plane is constructed, and in Section 5 the 166 detailed process is explained. Section 6 explains how non-autonomic 167 nodes and networks can be integrated, Section 7 defines the 168 negotiation protocol, and Section 7.3 the first channel types for the 169 ACP. 171 The document "Autonomic Network Stable Connectivity" 172 [I-D.eckert-anima-stable-connectivity] describes how the ACP can be 173 used to provide stable connectivity for OAM applications. It also 174 explains on how existing management solutions can leverage the ACP in 175 parallel with traditional management models, when to use the ACP 176 versus the data plane, how to integrate IPv4 based management, etc. 178 2. Use Cases for an Autonomic Control Plane 180 2.1. An Infrastructure for Autonomic Functions 182 Autonomic Functions need a stable infrastructure to run on, and all 183 autonomic functions should use the same infrastructure to minimise 184 the complexity of the network. This way, there is only need for a 185 single discovery mechanism, a single security mechanism, and other 186 processes that distributed functions require. 188 2.2. Secure Bootstrap over an Unconfigured Network 190 Today, bootstrapping a new device typically requires all devices 191 between a controlling node (such as an SDN controller) and the new 192 device to be completely and correctly addressed, configured and 193 secured. Therefore, bootstrapping a network happens in layers around 194 the controller. Without console access (for example through an out 195 of band network) it is not possible today to make devices securely 196 reachable before having configured the entire network between. 198 With the ACP, secure bootstrap of new devices can happen without 199 requiring any configuration on the network. A new device can 200 automatically be bootstrapped in a secure fashion and be deployed 201 with a domain certificate. This does not require any configuration 202 on intermediate nodes, because they can communicate through the ACP. 204 2.3. Data Plane Independent Permanent Reachability 206 Today, most critical control plane protocols and network management 207 protocols are running in the data plane (global routing table) of the 208 network. This leads to undesirable dependencies between control and 209 management plane on one side and the data plane on the other: Only if 210 the data plane is operational, will the other planes work as 211 expected. 213 Data plane connectivity can be affected by errors and faults, for 214 example certain AAA misconfigurations can lock an administrator out 215 of a device; routing or addressing issues can make a device 216 unreachable; shutting down interfaces over which a current management 217 session is running can lock an admin irreversibly out of the device. 218 Traditionally only console access can help recover from such issues. 220 Data plane dependencies also affect NOC/SDN controller applications: 221 Certain network changes are today hard to operate, because the change 222 itself may affect reachability of the devices. Examples are address 223 or mask changes, routing changes, or security policies. Today such 224 changes require precise hop-by-hop planning. 226 The ACP provides reachability that is largely independent of the data 227 plane, which allows control plane and management plane to operate 228 more robustly: 230 o For management plane protocols, the ACP provides the functionality 231 of a "Virtual-out-of-band (VooB) channel", by providing 232 connectivity to all devices regardless of their configuration or 233 global routing table. 235 o For control plane protocols, the ACP allows their operation even 236 when the data plane is temporarily faulty, or during transitional 237 events, such as routing changes, which may affect the control 238 plane at least temporarily. This is specifically important for 239 autonomic service agents, which could affect data plane 240 connectivity. 242 The document "Autonomic Network Stable Connectivity" 243 [I-D.eckert-anima-stable-connectivity] explains the use cases for the 244 ACP in significantly more detail and explains how the ACP can be used 245 in practical network operations. 247 3. Requirements 249 The Autonomic Control Plane has the following requirements: 251 1. The ACP SHOULD provide robust connectivity: As far as possible, 252 it should be independent of configured addressing, configuration 253 and routing. Requirements 2 and 3 build on this requirement, but 254 also have value on their own. 256 2. The ACP MUST have a separate address space from the data plane. 257 Reason: traceability, debug-ability, separation from data plane, 258 security (can block easily at edge). 260 3. The ACP MUST use autonomically managed address space. Reason: 261 easy bootstrap and setup ("autonomic"); robustness (admin can't 262 mess things up so easily). This document suggests to use ULA 263 addressing for this purpose. 265 4. The ACP MUST be generic. Usable by all the functions and 266 protocols of the AN infrastructure. It MUST NOT be tied to a 267 particular protocol. 269 5. The ACP MUST provide security: Messages coming through the ACP 270 MUST be authenticated to be from a trusted node, and SHOULD (very 271 strong SHOULD) be encrypted. 273 The default mode of operation of the ACP is hop-by-hop, because this 274 interaction can be built on IPv6 link local addressing, which is 275 autonomic, and has no dependency on configuration (requirement 1). 276 It may be necessary to have end-to-end connectivity in some cases, 277 for example to provide an end-to-end security association for some 278 protocols. This is possible, but then has a dependency on routable 279 address space. 281 4. Overview 283 The Autonomic Control Plane is constructed in the following way (for 284 details, see Section 5): 286 o An autonomic node creates a virtual routing and forwarding (VRF) 287 instance, or a similar virtual context. 289 o It determines, following a policy, a candidate peer list. This is 290 the list of nodes to which it should establish an autonomic 291 control plane. Default policy is: To all adjacent nodes in the 292 same domain. Intent can override this default policy. 294 o For each node in the candidate peer list, it authenticates that 295 node and negotiates a mutually acceptable channel type. 297 o It then establishes a secure tunnel of the negotiated channel 298 type. These tunnels are placed into the previously set up VRF. 299 This creates an overlay network with hop-by-hop tunnels. 301 o Inside the ACP VRF, each node sets up a virtual interface with its 302 ULA IPv6 address. 304 o Each node runs a lightweight routing protocol, to announce 305 reachability of the virtual addresses inside the ACP. 307 o Non-autonomic NMS systems or controllers have to be manually 308 connected into the ACP. 310 o Connecting over non-autonomic Layer-3 clouds initially requires a 311 tunnel between autonomic nodes. 313 o None of the above operations (except manual ones) is reflected in 314 the configuration of the device. 316 The following figure illustrates the ACP. 318 autonomic node 1 autonomic node 2 319 ................... ................... 320 secure . . secure . . secure 321 tunnel : +-----------+ : tunnel : +-----------+ : tunnel 322 ..--------| ACP VRF |---------------------| ACP VRF |---------.. 323 : / \ / \ <--routing--> / \ / \ : 324 : \ / \ / \ / \ / : 325 ..--------| virtual |---------------------| virtual |---------.. 326 : | interface | : : | interface | : 327 : +-----------+ : : +-----------+ : 328 : : : : 329 : data plane :...............: data plane : 330 : : link : : 331 :.................: :.................: 333 Figure 1 335 The resulting overlay network is normally based exclusively on hop- 336 by-hop tunnels. This is because addressing used on links is IPv6 337 link local addressing, which does not require any prior set-up. This 338 way the ACP can be built even if there is no configuration on the 339 devices, or if the data plane has issues such as addressing or 340 routing problems. 342 5. Self-Creation of an Autonomic Control Plane 344 This section describes the steps to set up an Autonomic Control 345 Plane, and highlights the key properties which make it 346 "indestructible" against many inadvert changes to the data plane, for 347 example caused by misconfigurations. 349 5.1. Preconditions 351 An autonomic node can be a router, switch, controller, NMS host, or 352 any other IP device. We assume an autonomic node has: 354 o A globally unique domain certificate, with which it can 355 cryptographically assert its membership of the domain. The 356 document [I-D.ietf-anima-bootstrapping-keyinfra] describes how a 357 domain certificate can be automatically and securely derived from 358 a vendor specific Unique Device Identifier (UDI) or IDevID 359 certificate. (Note the UDI used in this document is NOT the UUID 360 specified in [RFC4122].) 362 o An adjacency table, which contains information about adjacent 363 autonomic nodes, at a minimum: node-ID, IP address, domain, 364 certificate. An autonomic device maintains this adjacency table 365 up to date. Where the next autonomic device is not directly 366 adjacent, the information in the adjacency table can be 367 supplemented by configuration. For example, the node-ID and IP 368 address could be configured. 370 The adjacency table MAY contain information about the validity and 371 trust of the adjacent autonomic node's certificate. However, 372 subsequent steps MUST always start with authenticating the peer. 374 The adjacency table contains information about adjacent autonomic 375 nodes in general, independently of their domain and trust status. 376 The next step determines to which of those autonomic nodes an ACP 377 connection should be established. 379 5.2. Candidate ACP Neighbor Selection 381 An autonomic node must determine to which other autonomic nodes in 382 the adjacency table it should build an ACP connection. 384 The ACP is by default established exclusively between nodes in the 385 same domain. 387 Intent can change this default behaviour. The precise format for 388 this Intent needs to be defined outside this document. Example 389 Intent policies are: 391 o The ACP should be built between all sub-domains for a given parent 392 domain. For example: For domain "example.com", nodes of 393 "example.com", "access.example.com", "core.example.com" and 394 "city.core.example.com" should all establish one single ACP. 396 o Two domains should build one single ACP between themselves, for 397 example "example1.com" should establish the ACP also with nodes 398 from "example2.com". For this case, the two domains must be able 399 to validate their trust, typically by cross-signing their 400 certificate infrastructure. 402 The result of the candidate ACP neighbor selection process is a list 403 of adjacent or configured autonomic neighbors to which an ACP channel 404 should be established. The next step begins that channel 405 establishment. 407 5.3. Capability Negotiation 409 Autonomic devices may have different capabilities based on the type 410 of device, OS version, etc. To establish a trusted secure ACP 411 channel, devices must first negotiate their mutual capabilities in 412 the data plane. This allows for the support of different channel 413 types in the future. 415 For each node on the candidate ACP neighbor list, capabilities need 416 to be exchanged. The capability negotiation is based on GRASP 417 [I-D.ietf-anima-grasp]. The relevant protocol details are defined in 418 Section 7. This negotiation MUST be secure: The identity of the 419 other node MUST be validated during capability negotiation, and the 420 exchange MUST be authenticated. 422 The first parameter to be negotiated is the ACP Channel type. The 423 channel types are defined in Section 7.3. Other parameters may be 424 added later. 426 Intent may also influence the capability negotiation. For example, 427 Intent may require a minimum ACP tunnel security. This is outside 428 scope for this document. 430 5.4. Channel Establishment 432 After authentication and capability negotiation autonomic nodes 433 establish a secure channel towards the AN neighbors with the above 434 negotiated parameters. 436 The channel establishment MUST be authenticated. Whether or not, and 437 how, a channel is encrypted is part of the capability negotiation, 438 potentially controlled by Intent. 440 In order to be independent of configured link addresses, channels 441 SHOULD use IPv6 link local addresses between adjacent neighbors 442 wherever possible. This way, the ACP tunnels are independent of 443 correct network wide routing. 445 Since channels are by default established between adjacent neighbors, 446 the resulting overlay network does hop by hop encryption. Each node 447 decrypts incoming traffic from the ACP, and encrypts outgoing traffic 448 to its neighbors in the ACP. Routing is discussed in Section 5.7. 450 If two nodes are connected via several links, the ACP SHOULD be 451 established on every link, but it is possible to establish the ACP 452 only on a sub-set of links. Having an ACP channel on every link has 453 a number of advantages, for example it allows for a faster failover 454 in case of link failure, and it reflects the physical topology more 455 closely. Using a subset of links (for example, a single link), 456 reduces resource consumption on the devices, because state needs to 457 be kept per ACP channel. 459 5.5. Context Separation 461 The ACP is in a separate context from the normal data plane of the 462 device. This context includes the ACP channels IPv6 forwarding and 463 routing as well as any required higher layer ACP functions. 465 In classical network device platforms, a dedicated so called "Virtual 466 routing and forwarding instance" (VRF) is one logical implementation 467 option for the ACP. If possible by the platform SW architecture, 468 separation options that minimize shared components are preferred. 469 The context for the ACP needs to be established automatically during 470 bootstrap of a device. As much as possible it should be protected 471 from being modified unintentionally by data plane configuration. 473 Context separation improves security, because the ACP is not 474 reachable from the global routing table. Also, configuration errors 475 from the data plane setup do not affect the ACP. 477 [EDNOTE: Previous versions of this document also discussed an option 478 where the ACP runs in the data plane without logical separation. 479 Consensus is to focus only on the separated ACP now, and to remove 480 the ACP in the data plane from this document. See Appendix B for the 481 reasons for this decision.] 483 5.6. Addressing inside the ACP 485 The channels explained above typically only establish communication 486 between two adjacent nodes. In order for communication to happen 487 across multiple hops, the autonomic control plane requires internal 488 network wide valid addresses and routing. Each autonomic node must 489 create a virtual interface with a network wide unique address inside 490 the ACP context mentioned in Section 5.5. This address may be used 491 also in other virtual contexts. 493 With the algorithm introduced here, all autonomic devices in the same 494 domain have the same /48 prefix. Conversely, global IDs from 495 different domains are unlikely to clash, such that two networks can 496 be merged, as long as the policy allows that merge. See also 497 Section 8 for a discussion on merging domains. 499 Links inside the ACP only use link-local IPv6 addressing, such that 500 each node only requires one routable virtual address. 502 5.6.1. Fundamental Concepts of Autonomic Addressing 504 o Usage: Autonomic addresses are exclusively used for self- 505 management functions inside a trusted domain. They are not used 506 for user traffic. Communications with entities outside the 507 trusted domain use another address space, for example normally 508 managed routable address space. 510 o Separation: Autonomic address space is used separately from user 511 address space and other address realms. This supports the 512 robustness requirement. 514 o Loopback-only: Only loopback interfaces of autonomic nodes carry a 515 routable address; all other interfaces exclusively use IPv6 link 516 local for autonomic functions. The usage of IPv6 link local 517 addressing is discussed in [RFC7404]. 519 o Use-ULA: For loopback interfaces of autonomic nodes, we use Unique 520 Local Addresses (ULA), as specified in [RFC4193]. An alternative 521 scheme was discussed, using assigned ULA addressing. The 522 consensus was to use standard ULA, because it was deemed to be 523 sufficient. 525 o No external connectivity: They do not provide access to the 526 Internet. If a node requires further reaching connectivity, it 527 should use another, traditionally managed address scheme in 528 parallel. 530 The ACP is based exclusively on IPv6 addressing, for a variety of 531 reasons: 533 o Simplicity, reliability and scale: If other network layer 534 protocols were supported, each would have to have its own set of 535 security associations, routing table and process, etc. 537 o Autonomic functions do not require IPv4: Autonomic functions and 538 autonomic service agents are new concepts. They can be 539 exclusively built on IPv6 from day one. There is no need for 540 backward compatibility. 542 o OAM protocols no not require IPv4: The ACP may carry OAM 543 protocols. All relevant protocols (SNMP, TFTP, SSH, SCP, Radius, 544 Diameter, ...) are available in IPv6. 546 5.6.2. The Base Addressing Scheme 548 The Base ULA addressing scheme for autonomic nodes has the following 549 format: 551 8 40 3 77 552 +--+--------------+------+------------------------------------------+ 553 |FD| hash(domain) | Type | (sub-scheme) | 554 +--+--------------+------+------------------------------------------+ 556 Figure 2: Base Addressing Scheme 558 The first 48 bits follow the ULA scheme, as defined in [RFC4193], to 559 which a type field is added: 561 o "FD" identifies a locally defined ULA address. 563 o The "global ID" is set here to be a hash of the domain name, which 564 results in a pseudo-random 40 bit value. It is calculated as the 565 first 40 bits of the MD5 hash of the domain name, in the example 566 "example.com". 568 o Type: Set to 000 (3 zero bits). This field allows different 569 address sub-schemes in the future. The goal is to start with a 570 minimal number (ideally one) of sub-schemes initially, but to 571 allow for extensions later if and when required. This addresses 572 the "upgradability" requirement. Assignment of types for this 573 field should be maintained by IANA. 575 5.6.3. Possible Sub-Schemes 577 The sub-schemes listed here are not intended to be all supported 578 initially, but are listed for discussion. The final document should 579 define ideally only a single sub-scheme for now, and leave the other 580 "types" for later assignment. 582 5.6.3.1. Sub-Scheme 1 584 51 13 64 585 +------------------------+---------+--------------------------------+ 586 | (base scheme) | Zone ID | Device ID | 587 +------------------------+---------+--------------------------------+ 589 Figure 3: Addressing Scheme 1 591 The fields are defined as follows: [Editor's note: The lengths of the 592 fields is for discussion.] 594 o Zone ID: If set to all zero bits: The device ID bits are used as 595 an identifier (as opposed to a locator). This results in a non- 596 hierarchical, flat addressing scheme. Any other value indicates a 597 zone. See section Section 5.6.4 on how this field is used in 598 detail. 600 o Device ID: A unique value for each device. 602 The device ID is derived as follows: In an Autonomic Network, a 603 registrar is enrolling new devices. As part of the enrolment process 604 the registrar assigns a number to the device, which is unique for 605 this registrar, but not necessarily unique in the domain. The 64 bit 606 device ID is then composed as: 608 o 48 bit: Registrar ID, a number unique inside the domain that 609 identifies the registrar which assigned the name to the device. A 610 MAC address of the registrar can be used for this purpose. 612 o 16 bit: Device number, a number which is unique for a given 613 registrar, to identify the device. This can be a sequentially 614 assigned number. 616 The "device ID" itself is unique in a domain (i.e., the Zone-ID is 617 not required for uniqueness). Therefore, a device can be addressed 618 either as part of a flat hierarchy (zone ID = 0), or with an 619 aggregation scheme (any other zone ID). A address with zone-ID could 620 be interpreted as an identifier, with another zone-ID as a locator. 621 See Section 5.6.4 for a description of the zone bits. 623 5.6.3.2. Sub-Scheme 2 625 51 13 64-V ? 626 +------------------------+---------+----------------------------+---+ 627 | (base scheme) | Zone ID | Device ID | V | 628 +------------------------+---------+----------------------------+---+ 630 Figure 4: Addressing Scheme 2 632 The fields are defined as follows: [Editor's note: The lengths of the 633 fields is for discussion.] 635 o Zone ID: As in sub-scheme 1. 637 o Device ID: As in sub-scheme 1. 639 o V: Virtualization bit(s): 1 or more bits that indicate a virtual 640 context on an autonomic node. 642 In addition the scheme 1 (Section 5.6.3.1), this scheme allows the 643 direct addressing of specific virtual containers / VMs on an 644 autonomic node. An increasing number of hardware platforms have a 645 distributed architecture, with a base OS for the node itself, and the 646 support for hardware blades with potentially different OSs. The VMs 647 on the blades could be considered as separate autonomic nodes, in 648 which case it would make sense to be able to address them directly. 649 Autonomic Service Agents (ASAs) could be instantiated in either the 650 base OS, or one of the VMs on a blade. This addressing scheme allows 651 for the easy separation of the hardware context. 653 The location of the V bit(s) at the end of the address allows to 654 announce a single prefix for each autonomic node, while having 655 separate virtual contexts addressable directly. 657 5.6.4. Usage of the Zone Field 659 The "zone ID" allows for the introduction of structure in the 660 addressing scheme. 662 Zone = zero is the default addressing scheme in an autonomic domain. 663 Every autonomic node MUST respond to its ACP address with zone=0. 664 Used on its own this leads to a non-hierarchical address scheme, 665 which is suitable for networks up to a certain size. In this case, 666 the addresses primarily act as identifiers for the nodes, and 667 aggregation is not possible. 669 If aggregation is required, the 13 bit value allows for up to 8191 670 zones. The allocation of zone numbers may either happen 671 automatically through a to-be-defined algorithm; or it could be 672 configured and maintained manually. [We could divide the zone space 673 into manual and automatic space - to be discussed.] 675 If a device learns through an autonomic method or through 676 configuration that it is part of a zone, it MUST also respond to its 677 ACP address with that zone number. In this case the ACP loopback is 678 configured with two ACP addresses: One for zone 0 and one for the 679 assigned zone. This method allows for a smooth transition between a 680 flat addressing scheme and an hierarchical one. 682 (Theoretically, the 13 bits for the zone ID would allow also for two 683 levels of zones, introducing a sub-hierarchy. We do not think this 684 is required at this point, but a new type could be used in the future 685 to support such a scheme.) 687 Note: Another way to introduce hierarchy is to use sub-domains in the 688 naming scheme. The node names "node17.subdomainA.example.com" and 689 "node4.subdomainB.example.com" would automatically lead to different 690 ULA prefixes, which can be used to introduce a routing hierarchy in 691 the network, assuming that the subdomains are aligned with routing 692 areas. 694 5.7. Routing in the ACP 696 Once ULA address are set up all autonomic entities should run a 697 routing protocol within the autonomic control plane context. This 698 routing protocol distributes the ULA created in the previous section 699 for reachability. The use of the autonomic control plane specific 700 context eliminates the probable clash with the global routing table 701 and also secures the ACP from interference from the configuration 702 mismatch or incorrect routing updates. 704 The establishment of the routing plane and its parameters are 705 automatic and strictly within the confines of the autonomic control 706 plane. Therefore, no manual configuration is required. 708 All routing updates are automatically secured in transit as the 709 channels of the autonomic control plane are by default secured. 711 The routing protocol inside the ACP should be light weight and highly 712 scalable to ensure that the ACP does not become a limiting factor in 713 network scalability. We suggest the use of RPL [RFC6550] as one such 714 protocol which is light weight and scales well for the control plane 715 traffic. See Appendix A for more details on the choice of RPL. 717 6. Workarounds for Non-Autonomic Nodes 719 6.1. Connecting a Non-Autonomic Controller / NMS system 721 The Autonomic Control Plane can be used by management systems, such 722 as controllers or network management system (NMS) hosts (henceforth 723 called simply "NMS hosts"), to connect to devices through it. For 724 this, an NMS host must have access to the ACP. By default, the ACP 725 is a self-protecting overlay network, which only allows access to 726 trusted systems. Therefore, a traditional, non-autonomic NMS system 727 does not have access to the ACP by default, just like any other 728 external device. 730 If the NMS host is not autonomic, i.e., it does not support autonomic 731 negotiation of the ACP, then it can be brought into the ACP by 732 explicit configuration. On an adjacent autonomic node with ACP, the 733 interface with the NMS host can be configured to be part of the ACP. 734 In this case, the NMS host is with this interface entirely and 735 exclusively inside the ACP. It would likely require a second 736 interface for connections between the NMS host and administrators, or 737 Internet based services. This mode of connecting an NMS host has 738 security consequences: All systems and processes connected to this 739 implicitly trusted interface have access to all autonomic nodes on 740 the entire ACP, without further authentication. Thus, this 741 connection must be physically controlled. 743 The non-autonomic NMS host must be routed in the ACP. This involves 744 two parts: 1) the NMS host must point default to the AN device for 745 the ULA prefix used inside the ACP, and 2) the prefix used between AN 746 node and NMS host must be announced into the ACP, and distributed 747 there. 749 The document "Autonomic Network Stable Connectivity" 750 [I-D.eckert-anima-stable-connectivity] explains in more detail how 751 the ACP can be integrated in a mixed NOC environment. 753 6.2. ACP through Non-Autonomic L3 Clouds 755 Not all devices in a network may be autonomic. If non-autonomic 756 Layer-2 devices are between autonomic nodes, the communications 757 described in this document should work, since it is IP based. 758 However, non-autonomic Layer-3 devices do not forward link local 759 autonomic messages, and thus break the Autonomic Control Plane. 761 One workaround is to manually configure IP tunnels between autonomic 762 nodes across a non-autonomic Layer-3 cloud. The tunnels are 763 represented on each autonomic node as virtual interfaces, and all 764 autonomic transactions work across such tunnels. 766 Such manually configured tunnels are less "indestructible" than an 767 automatically created ACP based on link local addressing, since they 768 depend on correct data plane operations, such as routing and 769 addressing. 771 7. Building the ACP 773 7.1. Neighbor discovery via GRASP 775 Autonomic devices use inscure GRASP to discovery candidate autonomic 776 neighbors across L2 adjacencies. When Alice discovers Bob: 778 o If Alice is not part an autonomic domain, she starts autonomic 779 enrollment with Bob as the proxy using procedures described in 780 [I-D.ietf-anima-bootstrapping-keyinfra]. 782 o If Alice is part of an autonomic domain, Alice attempts to build 783 the ACP to Bob. Bob will do the same. 785 7.2. Channel Selection 787 To avoid attacks, initial discovery of candidate ACP peers can not 788 include any non-protected negotiation. To avoid re-inventing and 789 validating security association mechanisms, the next step after 790 discoving the address of a candidate neighbor can only be to try 791 first to establish a security association with that neighbor using a 792 well-known security association method. 794 At this time in the lifecycle of autonomic devices, it is unclear 795 whether it is feasible to even decide on a single MTI (mandatory to 796 implement) security association protocol across all autonomic 797 devices. 799 From the use-cases it is clear that not all type of autonomic devices 800 can or need to connect directly to each other or are able to support 801 or prefer all possible mechanisms. For example, code space limited 802 IoT devices may only support dTLS (because that code exists already 803 on them for end-to-end security use-cases), but low-end in-ceiling L2 804 switches may only want to support MacSec because that is also 805 supported in HW, and only a more flexible garteway device may need to 806 support both of these mechanisms and potentially more. 808 To support these requirements, and make ACP channel negotiation also 809 easily extensible, the secure channel selection between Alice and Bob 810 is a potentially two stage process. In the first stage, Alice and 811 Bob directly try to establish a secure channel using the security- 812 association and channel protocols they support. One or more of these 813 protocols may simply be protocols not directly resulting in an ACP 814 channel, but instead establishing a secure negotiation channel 815 through which the final secure channel protocol is decided. If both 816 Alice and Bob support such a negotiation step, then this secured 817 negotiation channel is the first step, and the secure channel 818 protocol is the second step. 820 If Alice supports multiple security association protocols in the 821 first step, it is a matter of Alices local policy to determine the 822 order in which she will try to build the connection to Bob. To 823 support multiple security association protocols, Alice must be able 824 to simultaneously act as a responder in parallel for all of them - so 825 that she can respond to any order in which Bob wants to prefer 826 building the security association. 828 When ACP setup between Alice and Bob results in the first secure 829 association to be established, the peer with the higher Device-ID in 830 the certificate will stop building new security associations. The 831 peer with the lower certificate Device-ID is now responsible to 832 continue building its most preferred security association and to tear 833 down all but that most preferred one - unless the secure association 834 is one of a negotation protocols whose rules superceed this. 836 All this negotiation is in the context of an "L2 interface". Alice 837 and Bob will build ACP connections to each other on every "L2 838 interface" that they both connect to. 840 7.3. Security Association protocols 842 The following sections define the security association protocols that 843 we consider to be important and feasible to specify in this document. 844 In all cases, the mutual authentication is done via the autonomic 845 domain certificate of the peer as follows - unless superceeded by 846 intent: 848 o The certificate is valid as proven by the security associations 849 protocol exchanges. 851 o The peers certificate is signed by the same CA as the devices 852 domain certificate. 854 o The peers OU (Organizational Unit) field in the certificates 855 Subject is the same as in the devices certificate. 857 7.3.1. ACP via IPsec 859 To run ACP via IPsec transport mode, no further IANA assignments/ 860 definitions are required. All autonomic devices suppoting IPsec MUST 861 support IPsec security setup via IKEv2, transpoort mode encapsulation 862 via the device and peer link-local IPv6 addresses and AES256 863 encryption. Further parameter options can be negotiated via IKEv2 or 864 via GRASP/TLS. 866 7.3.2. ACP via GRE/IPsec 868 In network devices it is often easier to provide virtual interfaces 869 on top of GRE encapsulation than natively on top of a simple IPsec 870 association. On those devices it may be necessary to run the ACP 871 secure channel on top of a GRE connection protected by the IPsec 872 association. The requirements for the IPsec association are the same 873 as described above, but instead of directly carrying the ACP IPv6 874 packets, the payload is an ACP IPv6 packet inside GREP/IPv6. 876 Note that without explicit negotiation (eg: via GRASP/TLS), this 877 method is incompatible to direct ACP via IPsec, so it must only be 878 used as an option during GRASP/TLS negotiation. 880 7.3.3. ACP via dTLS 882 To run ACP via UDP and dTLS v1.2 [RFC6347] an IANA assigned port 883 [TBD] is used. All autonomic devices supporting ACP via dTLS must 884 support AES256 encryption. 886 7.3.4. GRASP/TLS negotiation 888 To explicitly allow negotiation of the ACP channel protocol, GRASP 889 over a TLS connection using the GRASP_LISTEN_PORT and the devices and 890 peers link-local IPv6 address is used. When Alice and Bob support 891 GRASP negotiation, they do prefer it over any other non-explicitly 892 negotiated security association protocol and should wait trying any 893 non-negotiated ACP channel protocol until after it is clear that 894 GRASP/TLS will not work to the peer. 896 When Alice and Bob successfully establish the GRASP/TSL session, they 897 will initially negotiate the channel mechanism to use. Bob and Alice 898 each have a list of channel mehanisms they support, sorted by 899 preference. They negotiate the best mechansm supported by both of 900 them. In the absence of Intent, the mechanism choosen is the best 901 one for the device with the lower Device-ID. 903 After agreeing on a channel mechanism, Alice and Bob start the 904 selected Channel protocol. The GRASP/TLS connection can be kept 905 alive or timed out as long as the seelected channel protocol has a 906 secure association between Alice and Bob. When it terminates, it 907 needs to be re-negotiated via GRASP/TLS. 909 Negotiation of a channel type may require IANA assignments of code 910 points. See IANA Considerations (Section 13) for the formal 911 definition of those code points. 913 TBD: The exact negotiation steps in GRASP to achieve this outcome. 915 7.3.5. ACP Security Profiles 917 A baseline autonomic device MUST support IPsec and SHOULD support 918 GRASP/TLS and dTLS. A constrained autonomic device MUST support 919 dTLS. 921 Autonomic devices need to specify in documentation the set of secure 922 ACP mechanisms they suppport. 924 7.4. GRASP instance details 926 GRASP run to (insecurely) discover autonomic neighbors are isolated 927 instances from each other and other uses of GRASP - GRASP/TLS 928 sessions of L2 interfaces and GRASP inside the ACP 930 Received GRASP packets are assigned to an instance of GRASP by the 931 context they are received on: 933 o GRASP packets received on an ACP (virtual) interfaces are assigned 934 to the ACP instance of GRASP 936 o GRASP/UDP packets received on L2 interfaces where the device is 937 willing to run ACP across are are assigned to a separate instance 938 of GRASP for that L2 interface. We call those instances of GRASP 939 the "insecure L2 GRASP instances" and the ASA to perform the 940 discovery the "insecure L2 discovery ASA" (IL2D). 942 o GRASP packets received inside a TLS connection established for 943 GRASP/TLS ACP negotiation are assigned to a separate instance of 944 GRASP for that negotiation 946 All insecure L2 discovery of candidate ACP neighbors via GRASP and 947 the potentially following GRASP/TLS negotiation is per-L2 interface: 948 If Alice and Bob connect to each other via multiple interfaces, they 949 will independently establish the ACP to each other across each of 950 these interfaces. 952 For every L2-discovery instance of GRASP and its IL2D, the following 953 rules apply, amending and overriding the recommendations in 954 [I-D.ietf-anima-grasp]: 956 o GRASP link-local multicast discovery messages MUST use GTSM 957 [RFC5082]. With GTSM, discovery packets are sent with a TTL of 958 255 and packets received with a TTL smaller than 255 are ignored 959 upon receipt. 961 o The GRASP loop count of GRASP discovery packets must be set to 1 962 on sending. 964 o GRASP MUST send response messages for the discovery objected 965 defined here (overriding the MAY). 967 o GRASP MUST NOT respond to discovery objectives with the Divert 968 option - objectives learned and cached are solely for local 969 consumption. 971 o GRASP MUST NOT relay discovery or any other messages across 972 different interfaces. 974 TBD: The Details of the GRASP objective/packet formats still need to 975 be defined. Eg: Need to define an allocation for the objective of 976 "Autonomic Node". 978 8. Self-Healing Properties 980 The ACP is self-healing: 982 o New neighbors will automatically join the ACP after successful 983 validation and will become reachable using their unique ULA 984 address across the ACP. 986 o When any changes happen in the topology, the routing protocol used 987 in the ACP will automatically adapt to the changes and will 988 continue to provide reachability to all devices. 990 o If an existing device gets revoked, it will automatically be 991 denied access to the ACP as its domain certificate will be 992 validated against a Certificate Revocation List during 993 authentication. Since the revocation check is only done at the 994 establishment of a new security association, existing ones are not 995 automatically torn down. If an immediate disconnect is required, 996 existing sessions to a freshly revoked device can be re-set. 998 The ACP can also sustain network partitions and mergers. Practically 999 all ACP operations are link local, where a network partition has no 1000 impact. Devices authenticate each other using the domain 1001 certificates to establish the ACP locally. Addressing inside the ACP 1002 remains unchanged, and the routing protocol inside both parts of the 1003 ACP will lead to two working (although partitioned) ACPs. 1005 There are few central dependencies: A certificate revocation list 1006 (CRL) may not be available during a network partition; a suitable 1007 policy to not immediately disconnect neighbors when no CRL is 1008 available can address this issue. Also, a registrar or Certificate 1009 Authority might not be available during a partition. This may delay 1010 renewal of certificates that are to expire in the future, and it may 1011 prevent the enrolment of new devices during the partition. 1013 After a network partition, a re-merge will just establish the 1014 previous status, certificates can be renewed, the CRL is available, 1015 and new devices can be enrolled everywhere. Since all devices use 1016 the same trust anchor, a re-merge will be smooth. 1018 Merging two networks with different trust anchors requires the trust 1019 anchors to mutually trust each other (for example, by cross-signing). 1020 As long as the domain names are different, the addressing will not 1021 overlap (see Section 5.6). 1023 9. Self-Protection Properties 1025 As explained in Section 5, the ACP is based on channels being built 1026 between devices which have been previously authenticated based on 1027 their domain certificates. The channels themselves are protected 1028 using standard encryption technologies like DTLS or IPsec which 1029 provide additional authentication during channel establishment, data 1030 integrity and data confidentiality protection of data inside the ACP 1031 and in addition, provide replay protection. 1033 An attacker will therefore not be able to join the ACP unless having 1034 a valid domain certificate, also packet injection and sniffing 1035 traffic will not be possible due to the security provided by the 1036 encryption protocol. 1038 The remaining attack vector would be to attack the underlying AN 1039 protocols themselves, either via directed attacks or by denial-of- 1040 service attacks. However, as the ACP is built using link-local IPv6 1041 address, remote attacks are impossible. The ULA addresses are only 1042 reachable inside the ACP context, therefore unreachable from the data 1043 plane. Also, the ACP protocols should be implemented to be attack 1044 resistant and not consume unnecessary resources even while under 1045 attack. 1047 10. The Administrator View 1049 An ACP is self-forming, self-managing and self-protecting, therefore 1050 has minimal dependencies on the administrator of the network. 1051 Specifically, since it is independent of configuration, there is no 1052 scope for configuration errors on the ACP itself. The administrator 1053 may have the option to enable or disable the entire approach, but 1054 detailed configuration is not possible. This means that the ACP must 1055 not be reflected in the running configuration of devices, except a 1056 possible on/off switch. 1058 While configuration is not possible, an administrator must have full 1059 visibility of the ACP and all its parameters, to be able to do 1060 trouble-shooting. Therefore, an ACP must support all show and debug 1061 options, as for any other network function. Specifically, a network 1062 management system or controller must be able to discover the ACP, and 1063 monitor its health. This visibility of ACP operations must clearly 1064 be separated from visibility of data plane so automated systems will 1065 never have to deal with ACP aspect unless they explicitly desire to 1066 do so. 1068 Since an ACP is self-protecting, a device not supporting the ACP, or 1069 without a valid domain certificate cannot connect to it. This means 1070 that by default a traditional controller or network management system 1071 cannot connect to an ACP. See Section 6.1 for more details on how to 1072 connect an NMS host into the ACP. 1074 11. Explanations 1076 This section is non-normative and intended to provide further 1077 explanations for the choices made in this document. 1079 11.1. Why GRASP to discover autonomic neighbors 1081 None of the considered mechanisms to establish security associations 1082 (eg: IPsec or dTLS) include mechanisms to discover candidate 1083 neighbors, so these security mechanisms themselves are insufficient 1084 for the discovery. 1086 Existing L2 mechanisms such as LLDP (or vendor speccific alternatives 1087 like Ciscos CDP) are L2 link-local. If an autonomic device connects 1088 via an LLDP capable, but non-autonomic capable L2 switch to another 1089 autonomic device, then the non-autonomic L2 switch would not 1090 propagate the LLDP messages, so discovery would not work as desired. 1092 Existing L3/L4 link local discovery mechanisms such as mDNS or Web- 1093 Services Discovery (http://specs.xmlsoap.org/ws/2005/04/discovery/ws- 1094 discovery.pdf) are capable to support the simple discovery required 1095 by autonomic devices but have the following downsides compared to 1096 GRASP. 1098 There is no clear single ubiquitoous protocol that would apply 1099 equally well to all market segments in which autonomic routers are 1100 intended to be deployed. Making a choice is therefore difficult. 1102 In some of these protocols, the fact that they operate L3 link local 1103 is often seen as a limitation rather than as a necessity for the 1104 application. 1106 Various mechanisms are used or considered in these protocols to 1107 expand the scope of discovery beyond a single L3 subnet. If 1108 autonomic devices would use such a protocol, then autonomic discovery 1109 messages could more likely leak into remote networks and give more 1110 undesired (insecured) visibility into the presence of autonomic 1111 devices and potentially leading to more attempts to establish 1112 autonomic associations with those discovered devices. To achieve the 1113 maximum resilience with the minimum number of ACP channels, those 1114 channels need to follow as closely the physcial hops in the topology 1115 as possible. 1117 Visibility of discovery protocols in other domains may be 1118 undesirable: Visibility of mDNS messages for example could extend all 1119 the way into end user application level service browsers. It is 1120 undesirable to see desvices announcing themselves as automic there. 1122 Existing protocols can be more complex compared to GRASP as they have 1123 been designed for different purposes, for example to be more flexible 1124 and generic. In mDNS, if DNS-SD was used, it would require at least 1125 four RRs to be exchanged for a single service: a PTR, a SRV, a TXT 1126 and a AAAA RR. Minimizing the number of protocol exchanges by 1127 coalescing these RRs is possible but requires additional software 1128 design considerations. 1130 GRASP is already required inside the ACP and a highly desirable 1131 option for secure ACP channel negotiation (GRASP/TLS). Using it for 1132 discovery allows to reuse that already necessary code basis. If any 1133 other protocol was used for discovery, then autonomic discovery might 1134 be the only purpose for which the protocol code exists in the device. 1136 None of the above arguments individually are strong reasons not to 1137 use one of these GRASP alternatives, but together they make it 1138 reasonable to first define GRASP as the MTI (Mandatory To Implement) 1139 for the discovery step. 1141 12. Security Considerations 1143 An ACP is self-protecting and there is no need to apply configuration 1144 to make it secure. Its security therefore does not depend on 1145 configuration. 1147 However, the security of the ACP depends on a number of other 1148 factors: 1150 o The usage of domain certificates depends on a valid supporting PKI 1151 infrastructure. If the chain of trust of this PKI infrastructure 1152 is compromised, the security of the ACP is also compromised. This 1153 is typically under the control of the network administrator. 1155 o Security can be compromised by implementation errors (bugs), as in 1156 all products. 1158 Fundamentally, security depends on correct operation, implementation 1159 and architecture. Autonomic approaches such as the ACP largely 1160 eliminate the dependency on correct operation; implementation and 1161 architectural mistakes are still possible, as in all networking 1162 technologies. 1164 13. IANA Considerations 1166 Section 7.3.3 describes ACP over dTLS, which requires a well-known 1167 UDP port. We request IANA to assign this UDP port for 'ACP over 1168 dTLS'. 1170 Section 7.3.4 describes an option for the channel negotiation, the 1171 'ACP channel type'. We request IANA to create a registry for 'ACP 1172 channel type'. 1174 The ACP channel type is a 8-bit unsigned integer. This document only 1175 assigns the first value. 1177 Number | Channel Type | RFC 1178 ---------+-----------------------------------+------------ 1179 0 | GRE tunnel protected with | this document 1180 | IPsec transport mode | 1181 1-255 | reserved for future channel types | 1183 Section 5.6.2 describes a 'type' field in the base addressing scheme. 1184 We request IANA to create a registry for the 'ACP addressing scheme 1185 type'. The initial value and definition will be determined in a 1186 later version of this document. 1188 14. Acknowledgements 1190 This work originated from an Autonomic Networking project at Cisco 1191 Systems, which started in early 2010. Many people contributed to 1192 this project and the idea of the Autonomic Control Plane, amongst 1193 which (in alphabetical order): Ignas Bagdonas, Parag Bhide, Alex 1194 Clemm, Yves Hertoghs, Bruno Klauser, Max Pritikin, Ravi Kumar 1195 Vadapalli. 1197 Further input and suggestions were received from: Rene Struik, Brian 1198 Carpenter, Benoit Claise. 1200 15. Change log [RFC Editor: Please remove] 1202 15.1. Initial version 1204 First version of this document: 1205 [I-D.behringer-autonomic-control-plane] 1207 15.2. draft-behringer-anima-autonomic-control-plane-00 1209 Initial version of the anima document; only minor edits. 1211 15.3. draft-behringer-anima-autonomic-control-plane-01 1213 o Clarified that the ACP should be based on, and support only IPv6. 1215 o Clarified in intro that ACP is for both, between devices, as well 1216 as for access from a central entity, such as an NMS. 1218 o Added a section on how to connect an NMS system. 1220 o Clarified the hop-by-hop crypto nature of the ACP. 1222 o Added several references to GDNP as a candidate protocol. 1224 o Added a discussion on network split and merge. Although, this 1225 should probably go into the certificate management story longer 1226 term. 1228 15.4. draft-behringer-anima-autonomic-control-plane-02 1230 Addresses (numerous) comments from Brian Carpenter. See mailing list 1231 for details. The most important changes are: 1233 o Introduced a new section "overview", to ease the understanding of 1234 the approach. 1236 o Merged the previous "problem statement" and "use case" sections 1237 into a mostly re-written "use cases" section, since they were 1238 overlapping. 1240 o Clarified the relationship with draft-eckert-anima-stable- 1241 connectivity 1243 15.5. draft-behringer-anima-autonomic-control-plane-03 1245 o Took out requirement for IPv6 --> that's in the reference doc. 1247 o Added requirement section. 1249 o Changed focus: more focus on autonomic functions, not only virtual 1250 out of band. This goes a bit throughout the document, starting 1251 with a changed abstract and intro. 1253 15.6. draft-ietf-anima-autonomic-control-plane-00 1255 No changes; re-submitted as WG document. 1257 15.7. draft-ietf-anima-autonomic-control-plane-01 1259 o Added some paragraphs in addressing section on "why IPv6 only", to 1260 reflect the discussion on the list. 1262 o Moved the data-plane ACP out of the main document, into an 1263 appendix. The focus is now the virtually separated ACP, since it 1264 has significant advantages, and isn't much harder to do. 1266 o Changed the self-creation algorithm: Part of the initial steps go 1267 into the reference document. This document now assumes an 1268 adjacency table, and domain certificate. How those get onto the 1269 device is outside scope for this document. 1271 o Created a new section 6 "workarounds for non-autonomic nodes", and 1272 put the previous controller section (5.9) into this new section. 1273 Now, section 5 is "autonomic only", and section 6 explains what to 1274 do with non-autonomic stuff. Much cleaner now. 1276 o Added an appendix explaining the choice of RPL as a routing 1277 protocol. 1279 o Formalised the creation process a bit more. Now, we create a 1280 "candidate peer list" from the adjacency table, and form the ACP 1281 with those candidates. Also it explains now better that policy 1282 (Intent) can influence the peer selection. (section 4 and 5) 1284 o Introduce a section for the capability negotiation protocol 1285 (section 7). This needs to be worked out in more detail. This 1286 will likely be based on GRASP. 1288 o Introduce a new parameter: ACP tunnel type. And defines it in the 1289 IANA considerations section. Suggest GRE protected with IPSec 1290 transport mode as the default tunnel type. 1292 o Updated links, lots of small edits. 1294 15.8. draft-ietf-anima-autonomic-control-plane-02 1296 o Added explicitly text for the ACP channel negotiation. 1298 o Merged draft-behringer-anima-autonomic-addressing-02 into this 1299 document, as suggested by WG chairs. 1301 16. References 1303 [I-D.behringer-anima-autonomic-addressing] 1304 Behringer, M., "An Autonomic IPv6 Addressing Scheme", 1305 draft-behringer-anima-autonomic-addressing-02 (work in 1306 progress), October 2015. 1308 [I-D.behringer-anima-reference-model] 1309 Behringer, M., Carpenter, B., Eckert, T., Ciavaglia, L., 1310 Liu, B., Jeff, J., and J. Strassner, "A Reference Model 1311 for Autonomic Networking", draft-behringer-anima- 1312 reference-model-04 (work in progress), October 2015. 1314 [I-D.behringer-autonomic-control-plane] 1315 Behringer, M., Bjarnason, S., BL, B., and T. Eckert, "An 1316 Autonomic Control Plane", draft-behringer-autonomic- 1317 control-plane-00 (work in progress), June 2014. 1319 [I-D.eckert-anima-stable-connectivity] 1320 Eckert, T. and M. Behringer, "Using Autonomic Control 1321 Plane for Stable Connectivity of Network OAM", draft- 1322 eckert-anima-stable-connectivity-02 (work in progress), 1323 October 2015. 1325 [I-D.ietf-anima-bootstrapping-keyinfra] 1326 Pritikin, M., Richardson, M., Behringer, M., and S. 1327 Bjarnason, "Bootstrapping Key Infrastructures", draft- 1328 ietf-anima-bootstrapping-keyinfra-02 (work in progress), 1329 March 2016. 1331 [I-D.ietf-anima-grasp] 1332 Bormann, C., Carpenter, B., and B. Liu, "A Generic 1333 Autonomic Signaling Protocol (GRASP)", draft-ietf-anima- 1334 grasp-04 (work in progress), March 2016. 1336 [RFC4122] Leach, P., Mealling, M., and R. Salz, "A Universally 1337 Unique IDentifier (UUID) URN Namespace", RFC 4122, 1338 DOI 10.17487/RFC4122, July 2005, 1339 . 1341 [RFC4193] Hinden, R. and B. Haberman, "Unique Local IPv6 Unicast 1342 Addresses", RFC 4193, DOI 10.17487/RFC4193, October 2005, 1343 . 1345 [RFC5082] Gill, V., Heasley, J., Meyer, D., Savola, P., Ed., and C. 1346 Pignataro, "The Generalized TTL Security Mechanism 1347 (GTSM)", RFC 5082, DOI 10.17487/RFC5082, October 2007, 1348 . 1350 [RFC6347] Rescorla, E. and N. Modadugu, "Datagram Transport Layer 1351 Security Version 1.2", RFC 6347, DOI 10.17487/RFC6347, 1352 January 2012, . 1354 [RFC6550] Winter, T., Ed., Thubert, P., Ed., Brandt, A., Hui, J., 1355 Kelsey, R., Levis, P., Pister, K., Struik, R., Vasseur, 1356 JP., and R. Alexander, "RPL: IPv6 Routing Protocol for 1357 Low-Power and Lossy Networks", RFC 6550, 1358 DOI 10.17487/RFC6550, March 2012, 1359 . 1361 [RFC7404] Behringer, M. and E. Vyncke, "Using Only Link-Local 1362 Addressing inside an IPv6 Network", RFC 7404, 1363 DOI 10.17487/RFC7404, November 2014, 1364 . 1366 [RFC7575] Behringer, M., Pritikin, M., Bjarnason, S., Clemm, A., 1367 Carpenter, B., Jiang, S., and L. Ciavaglia, "Autonomic 1368 Networking: Definitions and Design Goals", RFC 7575, 1369 DOI 10.17487/RFC7575, June 2015, 1370 . 1372 [RFC7576] Jiang, S., Carpenter, B., and M. Behringer, "General Gap 1373 Analysis for Autonomic Networking", RFC 7576, 1374 DOI 10.17487/RFC7576, June 2015, 1375 . 1377 Appendix A. Background on the choice of routing protocol 1379 In a pre-standard implementation, the "IPv6 Routing Protocol for Low- 1380 Power and Lossy Networks (RPL, [RFC6550] was chosen. This 1381 Appendix explains the reasoning behind that decision. 1383 Requirements for routing in the ACP are: 1385 o Self-management: The ACP must build automatically, without human 1386 intervention. Therefore routing protocol must also work 1387 completely automatically. RPL is a simple, self-managing 1388 protocol, which does not require zones or areas; it is also self- 1389 configuring, since configuration is carried as part of the 1390 protocol (see Section 6.7.6 of [RFC6550]). 1392 o Scale: The ACP builds over an entire domain, which could be a 1393 large enterprise or service provider network. The routing 1394 protocol must therefore support domains of 100,000 nodes or more, 1395 ideally without the need for zoning or separation into areas. RPL 1396 has this scale property. This is based on extensive use of 1397 default routing. RPL also has other scalability improvements, 1398 such as selecting only a subset of peers instead of all possible 1399 ones, and trickle support for information synchronisation. 1401 o Low resource consumption: The ACP supports traditional network 1402 infrastructure, thus runs in addition to traditional protocols. 1403 The ACP, and specifically the routing protocol must have low 1404 resource consumption both in terms of memory and CPU requirements. 1405 Specifically, at edge nodes, where memory and CPU are scarce, 1406 consumption should be minimal. RPL builds a destination-oriented 1407 directed acyclic graph (DODAG), where the main resource 1408 consumption is at the root of the DODAG. The closer to the edge 1409 of the network, the less state needs to be maintained. This 1410 adapts nicely to the typical network design. Also, all changes 1411 below a common parent node are kept below that parent node. 1413 o Support for unstructured address space: In the Autonomic 1414 Networking Infrastructure, node addresses are identifiers, and may 1415 not be assigned in a topological way. Also, nodes may move 1416 topologically, without changing their address. Therefore, the 1417 routing protocol must support completely unstructured address 1418 space. RPL is specifically made for mobile ad-hoc networks, with 1419 no assumptions on topologically aligned addressing. 1421 o Modularity: To keep the initial implementation small, yet allow 1422 later for more complex methods, it is highly desirable that the 1423 routing protocol has a simple base functionality, but can import 1424 new functional modules if needed. RPL has this property with the 1425 concept of "objective function", which is a plugin to modify 1426 routing behaviour. 1428 o Extensibility: Since the Autonomic Networking Infrastructure is a 1429 new concept, it is likely that changes in the way of operation 1430 will happen over time. RPL allows for new objective functions to 1431 be introduced later, which allow changes to the way the routing 1432 protocol creates the DAGs. 1434 o Multi-topology support: It may become necessary in the future to 1435 support more than one DODAG for different purposes, using 1436 different objective functions. RPL allow for the creation of 1437 several parallel DODAGs, should this be required. This could be 1438 used to create different topologies to reach different roots. 1440 o No need for path optimisation: RPL does not necessarily compute 1441 the optimal path between any two nodes. However, the ACP does not 1442 require this today, since it carries mainly non-delay-sensitive 1443 feedback loops. It is possible that different optimisation 1444 schemes become necessary in the future, but RPL can be expanded 1445 (see point "Extensibility" above). 1447 Appendix B. Alternative: An ACP without Separation 1449 Section 5 explains how the ACP is constructed as a virtually 1450 separated overlay network. An alternative ACP design can be achieved 1451 without the VRFs. In this case, the autonomic virtual addresses are 1452 part of the data plane, and subject to routing, filtering, QoS, etc 1453 on the data plane. The secure tunnels are in this case used by 1454 traffic to and from the autonomic address space. They are still 1455 required to provide the authentication function for all autonomic 1456 packets. 1458 At IETF 93 in Prague, the suggestion was made to not advance with the 1459 data plane ACP, and only continue with the virtually separate ACP. 1460 The reason for this decision is that the contextual separation of the 1461 ACP provides a range of benefits (more robustness, less potential 1462 interactions with user configurations), while it is not much harder 1463 to achieve. 1465 This appendix serves to explain the decision; it will be removed in 1466 the next version of the draft. 1468 Authors' Addresses 1470 Michael H. Behringer (editor) 1471 Cisco Systems 1472 Building D, 45 Allee des Ormes 1473 Mougins 06250 1474 France 1476 Email: mbehring@cisco.com 1478 Steinthor Bjarnason 1479 Cisco Systems 1481 Email: sbjarnas@cisco.com 1482 Balaji BL 1483 Cisco Systems 1485 Email: blbalaji@cisco.com 1487 Toerless Eckert 1488 Cisco Systems 1490 Email: eckert@cisco.com