idnits 2.17.1 draft-ietf-anima-autonomic-control-plane-04.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- ** 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 255: '... 1. The ACP SHOULD provide robust c...' RFC 2119 keyword, line 260: '... 2. The ACP MUST have a separate ad...' RFC 2119 keyword, line 264: '... 3. The ACP MUST use autonomically ...' RFC 2119 keyword, line 269: '... 4. The ACP MUST be generic. Usabl...' RFC 2119 keyword, line 270: '...rastructure. It MUST NOT be tied to a...' (18 more instances...) Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (October 31, 2016) is 2732 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 669, but not defined == Unused Reference: 'RFC5082' is defined on line 1375, but no explicit reference was found in the text == Unused Reference: 'RFC6762' is defined on line 1402, but no explicit reference was found in the text == Unused Reference: 'RFC6763' is defined on line 1406, but no explicit reference was found in the text == Outdated reference: A later version (-45) exists of draft-ietf-anima-bootstrapping-keyinfra-04 == Outdated reference: A later version (-15) exists of draft-ietf-anima-grasp-08 == Outdated reference: A later version (-10) exists of draft-ietf-anima-reference-model-02 ** Downref: Normative reference to an Informational draft: draft-ietf-anima-reference-model (ref. 'I-D.ietf-anima-reference-model') == Outdated reference: A later version (-10) exists of draft-ietf-anima-stable-connectivity-01 ** Downref: Normative reference to an Informational draft: draft-ietf-anima-stable-connectivity (ref. 'I-D.ietf-anima-stable-connectivity') ** Downref: Normative reference to an Informational draft: draft-richardson-anima-6join-discovery (ref. 'I-D.richardson-anima-6join-discovery') ** 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 (~~), 9 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 Cisco Systems 4 Intended status: Standards Track T. Eckert 5 Expires: May 4, 2017 6 S. Bjarnason 7 Arbor Networks 8 October 31, 2016 10 An Autonomic Control Plane 11 draft-ietf-anima-autonomic-control-plane-04 13 Abstract 15 Autonomic functions need a control plane to communicate, which 16 depends on some addressing and routing. This Autonomic Control Plane 17 should ideally be self-managing, and as independent as possible of 18 configuration. This document defines an "Autonomic Control Plane", 19 with the primary use as a control plane for autonomic functions. It 20 also serves as a "virtual out of band channel" for OAM communications 21 over a network that is not configured, or mis-configured. 23 Status of This Memo 25 This Internet-Draft is submitted in full conformance with the 26 provisions of BCP 78 and BCP 79. 28 Internet-Drafts are working documents of the Internet Engineering 29 Task Force (IETF). Note that other groups may also distribute 30 working documents as Internet-Drafts. The list of current Internet- 31 Drafts is at http://datatracker.ietf.org/drafts/current/. 33 Internet-Drafts are draft documents valid for a maximum of six months 34 and may be updated, replaced, or obsoleted by other documents at any 35 time. It is inappropriate to use Internet-Drafts as reference 36 material or to cite them other than as "work in progress." 38 This Internet-Draft will expire on May 4, 2017. 40 Copyright Notice 42 Copyright (c) 2016 IETF Trust and the persons identified as the 43 document authors. All rights reserved. 45 This document is subject to BCP 78 and the IETF Trust's Legal 46 Provisions Relating to IETF Documents 47 (http://trustee.ietf.org/license-info) in effect on the date of 48 publication of this document. Please review these documents 49 carefully, as they describe your rights and restrictions with respect 50 to this document. Code Components extracted from this document must 51 include Simplified BSD License text as described in Section 4.e of 52 the Trust Legal Provisions and are provided without warranty as 53 described in the Simplified BSD License. 55 Table of Contents 57 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 58 2. Use Cases for an Autonomic Control Plane . . . . . . . . . . 4 59 2.1. An Infrastructure for Autonomic Functions . . . . . . . . 4 60 2.2. Secure Bootstrap over an Unconfigured Network . . . . . . 5 61 2.3. Data Plane Independent Permanent Reachability . . . . . . 5 62 3. Requirements . . . . . . . . . . . . . . . . . . . . . . . . 6 63 4. Overview . . . . . . . . . . . . . . . . . . . . . . . . . . 7 64 5. Self-Creation of an Autonomic Control Plane . . . . . . . . . 8 65 5.1. Preconditions . . . . . . . . . . . . . . . . . . . . . . 8 66 5.1.1. Domain Certificate with ANIMA information . . . . . . 8 67 5.1.2. AN Adjacency Table . . . . . . . . . . . . . . . . . 9 68 5.2. Neighbor discovery . . . . . . . . . . . . . . . . . . . 10 69 5.2.1. L2 topology considerations . . . . . . . . . . . . . 10 70 5.2.2. CDP/LLDP/mDNS considerations . . . . . . . . . . . . 11 71 5.2.3. Discovery with GRASP . . . . . . . . . . . . . . . . 11 72 5.2.4. Discovery and BRSKY . . . . . . . . . . . . . . . . . 12 73 5.3. Candidate ACP Neighbor Selection . . . . . . . . . . . . 12 74 5.4. Channel Selection . . . . . . . . . . . . . . . . . . . . 13 75 5.5. Security Association protocols . . . . . . . . . . . . . 14 76 5.5.1. ACP via IPsec . . . . . . . . . . . . . . . . . . . . 14 77 5.5.2. ACP via GRE/IPsec . . . . . . . . . . . . . . . . . . 15 78 5.5.3. ACP via dTLS . . . . . . . . . . . . . . . . . . . . 15 79 5.5.4. GRASP/TLS negotiation . . . . . . . . . . . . . . . . 15 80 5.5.5. ACP Security Profiles . . . . . . . . . . . . . . . . 16 81 5.6. GRASP instance details . . . . . . . . . . . . . . . . . 16 82 5.7. Context Separation . . . . . . . . . . . . . . . . . . . 16 83 5.8. Addressing inside the ACP . . . . . . . . . . . . . . . . 16 84 5.8.1. Fundamental Concepts of Autonomic Addressing . . . . 17 85 5.8.2. The ACP Addressing Base Scheme . . . . . . . . . . . 18 86 5.8.3. ACP Addressing Sub-Scheme . . . . . . . . . . . . . . 18 87 5.8.4. Usage of the Zone Field . . . . . . . . . . . . . . . 20 88 5.8.5. Other ACP Addressing Sub-Schemes . . . . . . . . . . 20 89 5.9. Routing in the ACP . . . . . . . . . . . . . . . . . . . 21 90 5.10. General ACP Considerations . . . . . . . . . . . . . . . 21 91 6. Workarounds for Non-Autonomic Nodes . . . . . . . . . . . . . 22 92 6.1. Connecting a Non-Autonomic Controller / NMS system . . . 22 93 6.2. ACP through Non-Autonomic L3 Clouds . . . . . . . . . . . 22 94 7. Self-Healing Properties . . . . . . . . . . . . . . . . . . . 23 95 8. Self-Protection Properties . . . . . . . . . . . . . . . . . 24 96 9. The Administrator View . . . . . . . . . . . . . . . . . . . 24 97 10. Security Considerations . . . . . . . . . . . . . . . . . . . 25 98 11. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 25 99 12. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 26 100 13. Change log [RFC Editor: Please remove] . . . . . . . . . . . 26 101 13.1. Initial version . . . . . . . . . . . . . . . . . . . . 26 102 13.2. draft-behringer-anima-autonomic-control-plane-00 . . . . 26 103 13.3. draft-behringer-anima-autonomic-control-plane-01 . . . . 26 104 13.4. draft-behringer-anima-autonomic-control-plane-02 . . . . 27 105 13.5. draft-behringer-anima-autonomic-control-plane-03 . . . . 27 106 13.6. draft-ietf-anima-autonomic-control-plane-00 . . . . . . 27 107 13.7. draft-ietf-anima-autonomic-control-plane-01 . . . . . . 27 108 13.8. draft-ietf-anima-autonomic-control-plane-02 . . . . . . 28 109 13.9. draft-ietf-anima-autonomic-control-plane-03 . . . . . . 28 110 13.10. draft-ietf-anima-autonomic-control-plane-04 . . . . . . 29 111 14. References . . . . . . . . . . . . . . . . . . . . . . . . . 29 112 Appendix A. Background on the choice of routing protocol . . . . 31 113 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 32 115 1. Introduction 117 Autonomic Networking is a concept of self-management: Autonomic 118 functions self-configure, and negotiate parameters and settings 119 across the network. [RFC7575] defines the fundamental ideas and 120 design goals of Autonomic Networking. A gap analysis of Autonomic 121 Networking is given in [RFC7576]. The reference architecture for 122 Autonomic Networking in the IETF is currently being defined in the 123 document [I-D.ietf-anima-reference-model] 125 Autonomic functions need a stable and robust infrastructure to 126 communicate on. This infrastructure should be as robust as possible, 127 and it should be re-usable by all autonomic functions. [RFC7575] 128 calls it the "Autonomic Control Plane". This document defines the 129 Autonomic Control Plane. 131 Today, the management and control plane of networks typically runs in 132 the global routing table, which is dependent on correct configuration 133 and routing. Misconfigurations or routing problems can therefore 134 disrupt management and control channels. Traditionally, an out of 135 band network has been used to recover from such problems, or 136 personnel is sent on site to access devices through console ports. 137 However, both options are operationally expensive. 139 In increasingly automated networks either controllers or distributed 140 autonomic service agents in the network require a control plane which 141 is independent of the network they manage, to avoid impacting their 142 own operations. 144 This document describes options for a self-forming, self-managing and 145 self-protecting "Autonomic Control Plane" (ACP) which is inband on 146 the network, yet as independent as possible of configuration, 147 addressing and routing problems (for details how this achieved, see 148 Section 5). It therefore remains operational even in the presence of 149 configuration errors, addressing or routing issues, or where policy 150 could inadvertently affect control plane connectivity. The Autonomic 151 Control Plane serves several purposes at the same time: 153 o Autonomic functions communicate over the ACP. The ACP therefore 154 supports directly Autonomic Networking functions, as described in 155 [I-D.ietf-anima-reference-model]. For example, GRASP 156 [I-D.ietf-anima-grasp] can run inside the ACP. 158 o An operator can use it to log into remote devices, even if the 159 data plane is misconfigured or unconfigured. 161 o A controller or network management system can use it to securely 162 bootstrap network devices in remote locations, even if the network 163 in between is not yet configured; no data-plane dependent 164 bootstrap configuration is required. An example of such a secure 165 bootstrap process is described in 166 [I-D.ietf-anima-bootstrapping-keyinfra] 168 This document describes some use cases for the ACP in Section 2, it 169 defines the requirements in Section 3, Section 4 gives an overview 170 how an Autonomic Control Plane is constructed, and in Section 5 the 171 detailed process is explained. Section 6 explains how non-autonomic 172 nodes and networks can be integrated, and Section 5.5 the first 173 channel types for the ACP. 175 The document "Autonomic Network Stable Connectivity" 176 [I-D.ietf-anima-stable-connectivity] describes how the ACP can be 177 used to provide stable connectivity for OAM applications. It also 178 explains on how existing management solutions can leverage the ACP in 179 parallel with traditional management models, when to use the ACP 180 versus the data plane, how to integrate IPv4 based management, etc. 182 2. Use Cases for an Autonomic Control Plane 184 2.1. An Infrastructure for Autonomic Functions 186 Autonomic Functions need a stable infrastructure to run on, and all 187 autonomic functions should use the same infrastructure to minimise 188 the complexity of the network. This way, there is only need for a 189 single discovery mechanism, a single security mechanism, and other 190 processes that distributed functions require. 192 2.2. Secure Bootstrap over an Unconfigured Network 194 Today, bootstrapping a new device typically requires all devices 195 between a controlling node (such as an SDN controller) and the new 196 device to be completely and correctly addressed, configured and 197 secured. Therefore, bootstrapping a network happens in layers around 198 the controller. Without console access (for example through an out 199 of band network) it is not possible today to make devices securely 200 reachable before having configured the entire network between. 202 With the ACP, secure bootstrap of new devices can happen without 203 requiring any configuration on the network. A new device can 204 automatically be bootstrapped in a secure fashion and be deployed 205 with a domain certificate. This does not require any configuration 206 on intermediate nodes, because they can communicate through the ACP. 208 2.3. Data Plane Independent Permanent Reachability 210 Today, most critical control plane protocols and network management 211 protocols are running in the data plane (global routing table) of the 212 network. This leads to undesirable dependencies between control and 213 management plane on one side and the data plane on the other: Only if 214 the data plane is operational, will the other planes work as 215 expected. 217 Data plane connectivity can be affected by errors and faults, for 218 example certain AAA misconfigurations can lock an administrator out 219 of a device; routing or addressing issues can make a device 220 unreachable; shutting down interfaces over which a current management 221 session is running can lock an admin irreversibly out of the device. 222 Traditionally only console access can help recover from such issues. 224 Data plane dependencies also affect NOC/SDN controller applications: 225 Certain network changes are today hard to operate, because the change 226 itself may affect reachability of the devices. Examples are address 227 or mask changes, routing changes, or security policies. Today such 228 changes require precise hop-by-hop planning. 230 The ACP provides reachability that is largely independent of the data 231 plane, which allows control plane and management plane to operate 232 more robustly: 234 o For management plane protocols, the ACP provides the functionality 235 of a "Virtual-out-of-band (VooB) channel", by providing 236 connectivity to all devices regardless of their configuration or 237 global routing table. 239 o For control plane protocols, the ACP allows their operation even 240 when the data plane is temporarily faulty, or during transitional 241 events, such as routing changes, which may affect the control 242 plane at least temporarily. This is specifically important for 243 autonomic service agents, which could affect data plane 244 connectivity. 246 The document "Autonomic Network Stable Connectivity" 247 [I-D.ietf-anima-stable-connectivity] explains the use cases for the 248 ACP in significantly more detail and explains how the ACP can be used 249 in practical network operations. 251 3. Requirements 253 The Autonomic Control Plane has the following requirements: 255 1. The ACP SHOULD provide robust connectivity: As far as possible, 256 it should be independent of configured addressing, configuration 257 and routing. Requirements 2 and 3 build on this requirement, but 258 also have value on their own. 260 2. The ACP MUST have a separate address space from the data plane. 261 Reason: traceability, debug-ability, separation from data plane, 262 security (can block easily at edge). 264 3. The ACP MUST use autonomically managed address space. Reason: 265 easy bootstrap and setup ("autonomic"); robustness (admin can't 266 mess things up so easily). This document suggests to use ULA 267 addressing for this purpose. 269 4. The ACP MUST be generic. Usable by all the functions and 270 protocols of the AN infrastructure. It MUST NOT be tied to a 271 particular protocol. 273 5. The ACP MUST provide security: Messages coming through the ACP 274 MUST be authenticated to be from a trusted node, and SHOULD (very 275 strong SHOULD) be encrypted. 277 The default mode of operation of the ACP is hop-by-hop, because this 278 interaction can be built on IPv6 link local addressing, which is 279 autonomic, and has no dependency on configuration (requirement 1). 280 It may be necessary to have end-to-end connectivity in some cases, 281 for example to provide an end-to-end security association for some 282 protocols. This is possible, but then has a dependency on routable 283 address space. 285 4. Overview 287 The Autonomic Control Plane is constructed in the following way (for 288 details, see Section 5): 290 o An autonomic node creates a virtual routing and forwarding (VRF) 291 instance, or a similar virtual context. 293 o It determines, following a policy, a candidate peer list. This is 294 the list of nodes to which it should establish an autonomic 295 control plane. Default policy is: To all adjacent nodes in the 296 same domain. Intent can override this default policy. 298 o For each node in the candidate peer list, it authenticates that 299 node and negotiates a mutually acceptable channel type. 301 o It then establishes a secure tunnel of the negotiated channel 302 type. These tunnels are placed into the previously set up VRF. 303 This creates an overlay network with hop-by-hop tunnels. 305 o Inside the ACP VRF, each node sets up a virtual interface with its 306 ULA IPv6 address. 308 o Each node runs a lightweight routing protocol, to announce 309 reachability of the virtual addresses inside the ACP. 311 o Non-autonomic NMS systems or controllers have to be manually 312 connected into the ACP. 314 o Connecting over non-autonomic Layer-3 clouds initially requires a 315 tunnel between autonomic nodes. 317 o None of the above operations (except manual ones) is reflected in 318 the configuration of the device. 320 The following figure illustrates the ACP. 322 autonomic node 1 autonomic node 2 323 ................... ................... 324 secure . . secure . . secure 325 tunnel : +-----------+ : tunnel : +-----------+ : tunnel 326 ..--------| ACP VRF |---------------------| ACP VRF |---------.. 327 : / \ / \ <--routing--> / \ / \ : 328 : \ / \ / \ / \ / : 329 ..--------| virtual |---------------------| virtual |---------.. 330 : | interface | : : | interface | : 331 : +-----------+ : : +-----------+ : 332 : : : : 333 : data plane :...............: data plane : 334 : : link : : 335 :.................: :.................: 337 Figure 1 339 The resulting overlay network is normally based exclusively on hop- 340 by-hop tunnels. This is because addressing used on links is IPv6 341 link local addressing, which does not require any prior set-up. This 342 way the ACP can be built even if there is no configuration on the 343 devices, or if the data plane has issues such as addressing or 344 routing problems. 346 5. Self-Creation of an Autonomic Control Plane 348 This section describes the steps to set up an Autonomic Control 349 Plane, and highlights the key properties which make it 350 "indestructible" against many inadvert changes to the data plane, for 351 example caused by misconfigurations. 353 5.1. Preconditions 355 An autonomic node can be a router, switch, controller, NMS host, or 356 any other IP device. We assume an autonomic node has a globally 357 unique domain certificate (LDevID), as well as an adjacency table. 359 5.1.1. Domain Certificate with ANIMA information 361 To establish an ACP securely, an Autnomic Node MUST have a globally 362 unique domain certificate (LDevID), with which it can 363 cryptographically assert its membership of the domain. The document 364 [I-D.ietf-anima-bootstrapping-keyinfra] describes how a domain 365 certificate can be automatically and securely derived from a vendor 366 specific Unique Device Identifier (UDI) or IDevID certificate. 367 (Note: the UDI used in this document is NOT the UUID specified in 368 [RFC4122].) 369 The domain certificate (LDevID) of an autonomic node MUST contain 370 ANIMA specific information, specifically the domain name, and its ACP 371 address with the zone-ID set to zero. This information MUST be 372 encoded in the LDevID in the subjectAltName / rfc822Name field in the 373 following way: 375 anima.acp+@ 377 An example: 379 anima.acp+FD99:B02D:8EC3:0:200:0:6400:1@example.com 381 The ACP address MUST be specified in its canonical form, as specified 382 in [RFC5952], to allow for easy textual comparisons. 384 The bootstrap process defined in 385 [I-D.ietf-anima-bootstrapping-keyinfra] MUST in an ANIMA network pass 386 on ACP address and domain to a new node, such that the new node can 387 add this to its enrolment request. 389 The Certificate Authority in an ANIMA network MUST honor these 390 parameters, and create the respective subjectAltName / rfc822Name in 391 the certificate. 393 ANIMA nodes can therefore find ACP address and domain using this 394 field in the domain certificate, both for themselves, as well as for 395 other nodes. 397 See section 4.2.1.6 of [RFC5280] for details on the subjectAltName 398 field. 400 5.1.2. AN Adjacency Table 402 To know to which nodes to establish an ACP channel, every autonomic 403 node maintains an adjacency table. The adjacency table contains 404 information about adjacent autonomic nodes, at a minimum: node-ID, IP 405 address, domain, certificate. An autonomic device MUST maintain this 406 adjacency table up to date. This table is used to determine to which 407 neighbor an ACP connection is established. 409 Where the next autonomic device is not directly adjacent, the 410 information in the adjacency table can be supplemented by 411 configuration. For example, the node-ID and IP address could be 412 configured. 414 The adjacency table MAY contain information about the validity and 415 trust of the adjacent autonomic node's certificate. However, 416 subsequent steps MUST always start with authenticating the peer. 418 The adjacency table contains information about adjacent autonomic 419 nodes in general, independently of their domain and trust status. 420 The next step determines to which of those autonomic nodes an ACP 421 connection should be established. 423 5.2. Neighbor discovery 425 5.2.1. L2 topology considerations 427 ANrtr1 ------ ANswitch1 --- ANswitch2 ------- ANrtr2 428 .../ \ \ ... 429 ANrtrI ------ \ ------- ANrtrN 430 ANswitchI ... 432 Figure 2 434 Consider a large L2 LAN with ANrtr1...ANrtrN connected via some 435 topology of L2 switches (eg: in a large enterprise campus or IoT 436 environment using large L2 LANs). If the discovery protocol used for 437 the ACP is operating at the subnet level, every AN router will see 438 all other AN routers on the LAN as neighbors and a full mesh of ACP 439 channels will be built. If some or all of the AN switches are 440 autonomic with the same discovery protocol, then the full mesh would 441 include those switches as well. 443 A full mesh of ACP connections like this can creates fundamental 444 challenges. The number of security associations of the secure 445 channel protocols will not scale arbitrarily, especially when they 446 leverage platform accelerated encryption/decryption. Likewise, any 447 other ACP operations needs to scale to the number of direct ACP 448 neigbors. An AN router with just 4 interfaces might be deployed into 449 a LAN with hundreds of neighbors connected via switches. Introducing 450 such a new unpredictable scaling factor requirement makes it harder 451 to support the ACP on arbitrary platforms and in arbitrary 452 deployments. 454 Predictable scaling requirements for ACP neighbors can most easily be 455 achieved if in topologies like these, AN capable L2 switches can 456 ensure that discovery messages terminate on them so that neighboring 457 AN routers and switches will only find the physcially connected AN L2 458 switches as their candidate ACP neighbors. With such a discovery 459 mechanism in place, the ACP and its security associations will only 460 need to scale to the number of physcial interfaces instead of a 461 potentially much larger number of "LAN-connected" neighbors. And the 462 ACP topology will follow directly the physical topology, something 463 which can then also be leveraged in management operations or by ASAs. 465 In the example above, consider ANswitch1 and ANswitchI are AN 466 capable, and ANswitch2 is not AN capable. The desired ACP topology 467 is therefore that ANrtr1 and ANrtrI only have an ACP connetion to 468 ANswitch1, and that ANswitch1, ANrtr2, ANrtrN have a full mesh of ACP 469 connection amongst each other. ANswitch1 also has an ACP connection 470 with ANswitchI and ANswitchI has ACP connections to anything else 471 behind it. 473 5.2.2. CDP/LLDP/mDNS considerations 475 LLDP (and Ciscos CDP) are example of L2 discovery protocols that do 476 terminate their messages on L2 ports. Unfortunately, they will also 477 terminate their messages if they do not support the ACP and would 478 then inhibit ACP neighbor discovery 480 mDNS operates at the subnet level, and is also used on L2 switches. 481 The authors of this document are not aware of mDNS implementation 482 that terminate their messages on L2 ports instead of the subnet 483 level. If mDNS was used as the ACP discovery mechanism on an ACP 484 capable L2 switch, then this would be necessary to implement. It is 485 likely that termination of mDNS messages could only be applied to all 486 mDNS messages from a port, which would then make it necessary to 487 software forward any non-ACP related mDNS messages to maintain prior 488 non-ACP mDNS functionality. With low performance of software 489 forwarding in many L2 switches, this could easily make the ACP 490 unsupportable on such L2 switches. 492 5.2.3. Discovery with GRASP 494 In conclusion for the above described considerations, the ACP uses 495 "insecure" instances of GRASP for discovery of ACP neighbors because 496 it can easily be set up to match the requiremetns without affecting 497 other uses of the protocol. 499 The name of the GRASP objective to announce/discover the capability 500 of a neighbor to run the ACP is "ACP". All other parameters are 501 defined in section [I-D.ietf-anima-grasp] where these instances of 502 GRASP are called "DULL" (Discovery Unsolicited Link Local). As 503 explained above, in an ACP enabled L2 switch, each of these instances 504 would actually need to be per-L2-port. The result of the discovery 505 is the IPv6 link-local address of the neighbor. It is stored in the 506 AN Adjacency Table, see Section 5.1.2 which then drives the further 507 building of the ACP to that neighbor. 509 For example, ANswitch1 would run separate DULL GRASP instances on its 510 ports to ANrtr1, ANswitch2 and ANswitchI, even though all those three 511 ports may be in the data plane in the same (V)LAN. This is easily 512 achieved by extracting native GRASP multicast messages by their MAC 513 multicast destination address. None of the other type of GRASP 514 instances (eg: as used inside the ACP) use GRASP messages that would 515 be affected by such extraction, because all other GRASP messages have 516 encrypted encapsulations. 518 5.2.4. Discovery and BRSKY 520 Before a node has a domain certificate, it can not participate in the 521 ACP and therefore does also not try to discover an ACP neighbor. 522 Instead, it uses the discovery mechanism described in 523 [I-D.ietf-anima-grasp] to discover a bootstrap proxy. Currently, 524 that document describes mDNS as the protocol of choice for that 525 discovery. In the context of above topology example, ANrtr1 might 526 therefore discover and choose any ANrtr or ANswitch on the LAN that 527 is already part of the autonomic domain - instead of the closest one 528 which is ANswitch1. This choice of bootstrap proxy does not impact 529 in the later building of the ACP on ANrtr1 and is therefore not a 530 concern for the ACP. 532 Once a device has its domain certificate, it will start announcing 533 itself via GRASP as ACP capable. 535 When an autonomic device is a registrar, it will announce the 536 registrar function via GRASP in the ACP as the "6JOIN" objective. An 537 AN device that is a registrar or learns about one or more reachable 538 registrars via this GRASP/ACP announcements will announce itself as a 539 boostrap proxy via mDNS. See [I-D.richardson-anima-6join-discovery] 540 for more details. 542 5.3. Candidate ACP Neighbor Selection 544 An autonomic node must determine to which other autonomic nodes in 545 the adjacency table it should build an ACP connection. This is based 546 on the information in the AN Adjacency table. 548 The ACP is by default established exclusively between nodes in the 549 same domain. 551 Intent can change this default behaviour. The precise format for 552 this Intent needs to be defined outside this document. Example 553 Intent policies are: 555 o The ACP should be built between all sub-domains for a given parent 556 domain. For example: For domain "example.com", nodes of 557 "example.com", "access.example.com", "core.example.com" and 558 "city.core.example.com" should all establish one single ACP. 560 o Two domains should build one single ACP between themselves, for 561 example "example1.com" should establish the ACP also with nodes 562 from "example2.com". For this case, the two domains must be able 563 to validate their trust, typically by cross-signing their 564 certificate infrastructure. 566 The result of the candidate ACP neighbor selection process is a list 567 of adjacent or configured autonomic neighbors to which an ACP channel 568 should be established. The next step begins that channel 569 establishment. 571 5.4. Channel Selection 573 To avoid attacks, initial discovery of candidate ACP peers can not 574 include any non-protected negotiation. To avoid re-inventing and 575 validating security association mechanisms, the next step after 576 discoving the address of a candidate neighbor can only be to try 577 first to establish a security association with that neighbor using a 578 well-known security association method. 580 At this time in the lifecycle of autonomic devices, it is unclear 581 whether it is feasible to even decide on a single MTI (mandatory to 582 implement) security association protocol across all autonomic 583 devices. 585 From the use-cases it is clear that not all type of autonomic devices 586 can or need to connect directly to each other or are able to support 587 or prefer all possible mechanisms. For example, code space limited 588 IoT devices may only support dTLS (because that code exists already 589 on them for end-to-end security use-cases), but low-end in-ceiling L2 590 switches may only want to support MacSec because that is also 591 supported in HW, and only a more flexible garteway device may need to 592 support both of these mechanisms and potentially more. 594 To support these requirements, and make ACP channel negotiation also 595 easily extensible, the secure channel selection between Alice and Bob 596 is a potentially two stage process. In the first stage, Alice and 597 Bob directly try to establish a secure channel using the security- 598 association and channel protocols they support. One or more of these 599 protocols may simply be protocols not directly resulting in an ACP 600 channel, but instead establishing a secure negotiation channel 601 through which the final secure channel protocol is decided. If both 602 Alice and Bob support such a negotiation step, then this secured 603 negotiation channel is the first step, and the secure channel 604 protocol is the second step. 606 If Alice supports multiple security association protocols in the 607 first step, it is a matter of Alices local policy to determine the 608 order in which she will try to build the connection to Bob. To 609 support multiple security association protocols, Alice must be able 610 to simultaneously act as a responder in parallel for all of them - so 611 that she can respond to any order in which Bob wants to prefer 612 building the security association. 614 When ACP setup between Alice and Bob results in the first secure 615 association to be established, the peer with the higher Device-ID in 616 the certificate will stop building new security associations. The 617 peer with the lower certificate Device-ID is now responsible to 618 continue building its most preferred security association and to tear 619 down all but that most preferred one - unless the secure association 620 is one of a negotation protocols whose rules superceed this. 622 All this negotiation is in the context of an "L2 interface". Alice 623 and Bob will build ACP connections to each other on every "L2 624 interface" that they both connect to. 626 5.5. Security Association protocols 628 The following sections define the security association protocols that 629 we consider to be important and feasible to specify in this document. 630 In all cases, the mutual authentication is done via the autonomic 631 domain certificate of the peer as follows - unless superceeded by 632 intent: 634 o The certificate is valid as proven by the security associations 635 protocol exchanges. 637 o The peers certificate is signed by the same CA as the devices 638 domain certificate. 640 o The peers OU (Organizational Unit) field in the certificates 641 Subject is the same as in the devices certificate. 643 5.5.1. ACP via IPsec 645 To run ACP via IPsec transport mode, no further IANA assignments/ 646 definitions are required. All autonomic devices suppoting IPsec MUST 647 support IPsec security setup via IKEv2, transport mode encapsulation 648 via the device and peer link-local IPv6 addresses and AES256 649 encryption. Further parameter options can be negotiated via IKEv2 or 650 via GRASP/TLS. 652 5.5.2. ACP via GRE/IPsec 654 In network devices it is often easier to provide virtual interfaces 655 on top of GRE encapsulation than natively on top of a simple IPsec 656 association. On those devices it may be necessary to run the ACP 657 secure channel on top of a GRE connection protected by the IPsec 658 association. The requirements for the IPsec association are the same 659 as described above, but instead of directly carrying the ACP IPv6 660 packets, the payload is an ACP IPv6 packet inside GRE/IPv6. 662 Note that without explicit negotiation (eg: via GRASP/TLS), this 663 method is incompatible to direct ACP via IPsec, so it must only be 664 used as an option during GRASP/TLS negotiation. 666 5.5.3. ACP via dTLS 668 To run ACP via UDP and dTLS v1.2 [RFC6347] an IANA assigned port 669 [TBD] is used. All autonomic devices supporting ACP via dTLS must 670 support AES256 encryption. 672 5.5.4. GRASP/TLS negotiation 674 To explicitly allow negotiation of the ACP channel protocol, GRASP 675 over a TLS connection using the GRASP_LISTEN_PORT and the devices and 676 peers link-local IPv6 address is used. When Alice and Bob support 677 GRASP negotiation, they do prefer it over any other non-explicitly 678 negotiated security association protocol and should wait trying any 679 non-negotiated ACP channel protocol until after it is clear that 680 GRASP/TLS will not work to the peer. 682 When Alice and Bob successfully establish the GRASP/TSL session, they 683 will initially negotiate the channel mechanism to use. Bob and Alice 684 each have a list of channel mehanisms they support, sorted by 685 preference. They negotiate the best mechansm supported by both of 686 them. In the absence of Intent, the mechanism choosen is the best 687 one for the device with the lower Device-ID. 689 After agreeing on a channel mechanism, Alice and Bob start the 690 selected Channel protocol. The GRASP/TLS connection can be kept 691 alive or timed out as long as the seelected channel protocol has a 692 secure association between Alice and Bob. When it terminates, it 693 needs to be re-negotiated via GRASP/TLS. 695 Negotiation of a channel type may require IANA assignments of code 696 points. See IANA Considerations (Section 11) for the formal 697 definition of those code points. 699 TBD: The exact negotiation steps in GRASP to achieve this outcome. 701 5.5.5. ACP Security Profiles 703 A baseline autonomic device MUST support IPsec and SHOULD support 704 GRASP/TLS and dTLS. A constrained autonomic device MUST support 705 dTLS. 707 Autonomic devices need to specify in documentation the set of secure 708 ACP mechanisms they suppport. 710 5.6. GRASP instance details 712 Received GRASP packets are assigned to an instance of GRASP by the 713 context they are received on: 715 o GRASP packets received on an ACP (virtual) interfaces are assigned 716 to the ACP instance of GRASP 718 o GRASP/UDP packets received on L2 interfaces/ports where the device 719 is willing to run ACP are assigned to a DULL instance of GRASP for 720 that interface/port. 722 o GRASP packets received inside a TLS connection established for 723 GRASP/TLS ACP negotiation are assigned to a separate instance of 724 GRASP for that negotiation. 726 5.7. Context Separation 728 The ACP is in a separate context from the normal data plane of the 729 device. This context includes the ACP channels IPv6 forwarding and 730 routing as well as any required higher layer ACP functions. 732 In classical network device platforms, a dedicated so called "Virtual 733 routing and forwarding instance" (VRF) is one logical implementation 734 option for the ACP. If possible by the platform SW architecture, 735 separation options that minimize shared components are preferred. 736 The context for the ACP needs to be established automatically during 737 bootstrap of a device. As much as possible it should be protected 738 from being modified unintentionally by data plane configuration. 740 Context separation improves security, because the ACP is not 741 reachable from the global routing table. Also, configuration errors 742 from the data plane setup do not affect the ACP. 744 5.8. Addressing inside the ACP 746 The channels explained above typically only establish communication 747 between two adjacent nodes. In order for communication to happen 748 across multiple hops, the autonomic control plane requires internal 749 network wide valid addresses and routing. Each autonomic node must 750 create a virtual interface with a network wide unique address inside 751 the ACP context mentioned in Section 5.7. This address may be used 752 also in other virtual contexts. 754 With the algorithm introduced here, all autonomic devices in the same 755 domain have the same /48 prefix. Conversely, global IDs from 756 different domains are unlikely to clash, such that two networks can 757 be merged, as long as the policy allows that merge. See also 758 Section 7 for a discussion on merging domains. 760 Links inside the ACP only use link-local IPv6 addressing, such that 761 each node only requires one routable virtual address. 763 5.8.1. Fundamental Concepts of Autonomic Addressing 765 o Usage: Autonomic addresses are exclusively used for self- 766 management functions inside a trusted domain. They are not used 767 for user traffic. Communications with entities outside the 768 trusted domain use another address space, for example normally 769 managed routable address space. 771 o Separation: Autonomic address space is used separately from user 772 address space and other address realms. This supports the 773 robustness requirement. 775 o Loopback-only: Only loopback interfaces of autonomic nodes carry a 776 routable address; all other interfaces exclusively use IPv6 link 777 local for autonomic functions. The usage of IPv6 link local 778 addressing is discussed in [RFC7404]. 780 o Use-ULA: For loopback interfaces of autonomic nodes, we use Unique 781 Local Addresses (ULA), as specified in [RFC4193]. An alternative 782 scheme was discussed, using assigned ULA addressing. The 783 consensus was to use standard ULA, because it was deemed to be 784 sufficient. 786 o No external connectivity: They do not provide access to the 787 Internet. If a node requires further reaching connectivity, it 788 should use another, traditionally managed address scheme in 789 parallel. 791 The ACP is based exclusively on IPv6 addressing, for a variety of 792 reasons: 794 o Simplicity, reliability and scale: If other network layer 795 protocols were supported, each would have to have its own set of 796 security associations, routing table and process, etc. 798 o Autonomic functions do not require IPv4: Autonomic functions and 799 autonomic service agents are new concepts. They can be 800 exclusively built on IPv6 from day one. There is no need for 801 backward compatibility. 803 o OAM protocols no not require IPv4: The ACP may carry OAM 804 protocols. All relevant protocols (SNMP, TFTP, SSH, SCP, Radius, 805 Diameter, ...) are available in IPv6. 807 5.8.2. The ACP Addressing Base Scheme 809 The Base ULA addressing scheme for autonomic nodes has the following 810 format: 812 8 40 3 77 813 +--+--------------+------+------------------------------------------+ 814 |FD| hash(domain) | Type | (sub-scheme) | 815 +--+--------------+------+------------------------------------------+ 817 Figure 3: ACP Addressing Base Scheme 819 The first 48 bits follow the ULA scheme, as defined in [RFC4193], to 820 which a type field is added: 822 o "FD" identifies a locally defined ULA address. 824 o The ULA "global ID" is set here to be a hash of the domain name, 825 which results in a pseudo-random 40 bit value. It is calculated 826 as the first 40 bits of the MD5 hash of the domain name, in the 827 example "example.com". 829 o Type: This field allows different address sub-schemes in the 830 future. The goal is to start with a single sub-schemes, but to 831 allow for extensions later if and when required. This addresses 832 the "upgradability" requirement. Assignment of types for this 833 field should be maintained by IANA. 835 5.8.3. ACP Addressing Sub-Scheme 837 The sub-scheme defined here is defined by the Type value 0 (zero) in 838 the base scheme. 840 51 13 63 1 841 +------------------------+---------+----------------------------+---+ 842 | (base scheme) | Zone ID | Device ID | V | 843 +------------------------+---------+----------------------------+---+ 845 Figure 4: ACP Addressing Sub-Scheme 847 The fields are defined as follows: [Editor's note: The lengths of the 848 fields is for discussion.] 850 o Zone ID: If set to all zero bits: The device ID bits are used as 851 an identifier (as opposed to a locator). This results in a non- 852 hierarchical, flat addressing scheme. Any other value indicates a 853 zone. See section Section 5.8.4 on how this field is used in 854 detail. 856 o Device ID: A unique value for each device. 858 o V: Virtualization bit: 0: autonomic node base system; 1: a virtual 859 context on an autonomic node. 861 The device ID is derived as follows: In an Autonomic Network, a 862 registrar is enrolling new devices. As part of the enrolment process 863 the registrar assigns a number to the device, which is unique for 864 this registrar, but not necessarily unique in the domain. The 64 bit 865 device ID is then composed as: 867 o 48 bit: Registrar ID, a number unique inside the domain that 868 identifies the registrar which assigned the name to the device. A 869 MAC address of the registrar can be used for this purpose. 871 o 15 bit: Device number, a number which is unique for a given 872 registrar, to identify the device. This can be a sequentially 873 assigned number. 875 The "device ID" itself is unique in a domain (i.e., the Zone-ID is 876 not required for uniqueness). Therefore, a device can be addressed 877 either as part of a flat hierarchy (zone ID = 0), or with an 878 aggregation scheme (any other zone ID). A address with zone-ID = 0 879 is an identifier, with another zone-ID as a locator. See 880 Section 5.8.4 for a description of the zone bits. 882 This addressing sub-scheme allows the direct addressing of specific 883 virtual containers / VMs on an autonomic node. An increasing number 884 of hardware platforms have a distributed architecture, with a base OS 885 for the node itself, and the support for hardware blades with 886 potentially different OSs. The VMs on the blades could be considered 887 as separate autonomic nodes, in which case it would make sense to be 888 able to address them directly. Autonomic Service Agents (ASAs) could 889 be instantiated in either the base OS, or one of the VMs on a blade. 890 This addressing scheme allows for the easy separation of the hardware 891 context. 893 The location of the V bit(s) at the end of the address allows to 894 announce a single prefix for each autonomic node, while having 895 separate virtual contexts addressable directly. 897 5.8.4. Usage of the Zone Field 899 The "zone ID" allows for the introduction of structure in the 900 addressing scheme. 902 Zone = zero is the default addressing scheme in an autonomic domain. 903 Every autonomic node MUST respond to its ACP address with zone=0. 904 Used on its own this leads to a non-hierarchical address scheme, 905 which is suitable for networks up to a certain size. In this case, 906 the addresses primarily act as identifiers for the nodes, and 907 aggregation is not possible. 909 If aggregation is required, the 13 bit value allows for up to 8191 910 zones. The allocation of zone numbers may either happen 911 automatically through a to-be-defined algorithm; or it could be 912 configured and maintained manually. [We could divide the zone space 913 into manual and automatic space - to be discussed.] 915 If a device learns through an autonomic method or through 916 configuration that it is part of a zone, it MUST also respond to its 917 ACP address with that zone number. In this case the ACP loopback is 918 configured with two ACP addresses: One for zone 0 and one for the 919 assigned zone. This method allows for a smooth transition between a 920 flat addressing scheme and an hierarchical one. 922 (Theoretically, the 13 bits for the zone ID would allow also for two 923 levels of zones, introducing a sub-hierarchy. We do not think this 924 is required at this point, but a new type could be used in the future 925 to support such a scheme.) 927 Note: Another way to introduce hierarchy is to use sub-domains in the 928 naming scheme. The node names "node17.subdomainA.example.com" and 929 "node4.subdomainB.example.com" would automatically lead to different 930 ULA prefixes, which can be used to introduce a routing hierarchy in 931 the network, assuming that the subdomains are aligned with routing 932 areas. 934 5.8.5. Other ACP Addressing Sub-Schemes 936 Other ACP addressing sub-schemes can be defined if and when required. 937 IANA will assign a new "type" for each new addressing sub-scheme. 939 5.9. Routing in the ACP 941 Once ULA address are set up all autonomic entities should run a 942 routing protocol within the autonomic control plane context. This 943 routing protocol distributes the ULA created in the previous section 944 for reachability. The use of the autonomic control plane specific 945 context eliminates the probable clash with the global routing table 946 and also secures the ACP from interference from the configuration 947 mismatch or incorrect routing updates. 949 The establishment of the routing plane and its parameters are 950 automatic and strictly within the confines of the autonomic control 951 plane. Therefore, no manual configuration is required. 953 All routing updates are automatically secured in transit as the 954 channels of the autonomic control plane are by default secured. 956 The routing protocol inside the ACP should be light weight and highly 957 scalable to ensure that the ACP does not become a limiting factor in 958 network scalability. We suggest the use of RPL [RFC6550] as one such 959 protocol which is light weight and scales well for the control plane 960 traffic. See Appendix A for more details on the choice of RPL. 962 5.10. General ACP Considerations 964 In order to be independent of configured link addresses, channels 965 SHOULD use IPv6 link local addresses between adjacent neighbors 966 wherever possible. This way, the ACP tunnels are independent of 967 correct network wide routing. 969 Since channels are by default established between adjacent neighbors, 970 the resulting overlay network does hop by hop encryption. Each node 971 decrypts incoming traffic from the ACP, and encrypts outgoing traffic 972 to its neighbors in the ACP. Routing is discussed in Section 5.9. 974 If two nodes are connected via several links, the ACP SHOULD be 975 established on every link, but it is possible to establish the ACP 976 only on a sub-set of links. Having an ACP channel on every link has 977 a number of advantages, for example it allows for a faster failover 978 in case of link failure, and it reflects the physical topology more 979 closely. Using a subset of links (for example, a single link), 980 reduces resource consumption on the devices, because state needs to 981 be kept per ACP channel. 983 6. Workarounds for Non-Autonomic Nodes 985 6.1. Connecting a Non-Autonomic Controller / NMS system 987 The Autonomic Control Plane can be used by management systems, such 988 as controllers or network management system (NMS) hosts (henceforth 989 called simply "NMS hosts"), to connect to devices through it. For 990 this, an NMS host must have access to the ACP. By default, the ACP 991 is a self-protecting overlay network, which only allows access to 992 trusted systems. Therefore, a traditional, non-autonomic NMS system 993 does not have access to the ACP by default, just like any other 994 external device. 996 If the NMS host is not autonomic, i.e., it does not support autonomic 997 negotiation of the ACP, then it can be brought into the ACP by 998 explicit configuration. On an adjacent autonomic node with ACP, the 999 interface with the NMS host can be configured to be part of the ACP. 1000 In this case, the NMS host is with this interface entirely and 1001 exclusively inside the ACP. It would likely require a second 1002 interface for connections between the NMS host and administrators, or 1003 Internet based services. This mode of connecting an NMS host has 1004 security consequences: All systems and processes connected to this 1005 implicitly trusted interface have access to all autonomic nodes on 1006 the entire ACP, without further authentication. Thus, this 1007 connection must be physically controlled. 1009 The non-autonomic NMS host must be routed in the ACP. This involves 1010 two parts: 1) the NMS host must point default to the AN device for 1011 the ULA prefix used inside the ACP, and 2) the prefix used between AN 1012 node and NMS host must be announced into the ACP, and distributed 1013 there. 1015 The document "Autonomic Network Stable Connectivity" 1016 [I-D.ietf-anima-stable-connectivity] explains in more detail how the 1017 ACP can be integrated in a mixed NOC environment. 1019 6.2. ACP through Non-Autonomic L3 Clouds 1021 Not all devices in a network may be autonomic. If non-autonomic 1022 Layer-2 devices are between autonomic nodes, the communications 1023 described in this document should work, since it is IP based. 1024 However, non-autonomic Layer-3 devices do not forward link local 1025 autonomic messages, and thus break the Autonomic Control Plane. 1027 One workaround is to manually configure IP tunnels between autonomic 1028 nodes across a non-autonomic Layer-3 cloud. The tunnels are 1029 represented on each autonomic node as virtual interfaces, and all 1030 autonomic transactions work across such tunnels. 1032 Such manually configured tunnels are less "indestructible" than an 1033 automatically created ACP based on link local addressing, since they 1034 depend on correct data plane operations, such as routing and 1035 addressing. 1037 7. Self-Healing Properties 1039 The ACP is self-healing: 1041 o New neighbors will automatically join the ACP after successful 1042 validation and will become reachable using their unique ULA 1043 address across the ACP. 1045 o When any changes happen in the topology, the routing protocol used 1046 in the ACP will automatically adapt to the changes and will 1047 continue to provide reachability to all devices. 1049 o If an existing device gets revoked, it will automatically be 1050 denied access to the ACP as its domain certificate will be 1051 validated against a Certificate Revocation List during 1052 authentication. Since the revocation check is only done at the 1053 establishment of a new security association, existing ones are not 1054 automatically torn down. If an immediate disconnect is required, 1055 existing sessions to a freshly revoked device can be re-set. 1057 The ACP can also sustain network partitions and mergers. Practically 1058 all ACP operations are link local, where a network partition has no 1059 impact. Devices authenticate each other using the domain 1060 certificates to establish the ACP locally. Addressing inside the ACP 1061 remains unchanged, and the routing protocol inside both parts of the 1062 ACP will lead to two working (although partitioned) ACPs. 1064 There are few central dependencies: A certificate revocation list 1065 (CRL) may not be available during a network partition; a suitable 1066 policy to not immediately disconnect neighbors when no CRL is 1067 available can address this issue. Also, a registrar or Certificate 1068 Authority might not be available during a partition. This may delay 1069 renewal of certificates that are to expire in the future, and it may 1070 prevent the enrolment of new devices during the partition. 1072 After a network partition, a re-merge will just establish the 1073 previous status, certificates can be renewed, the CRL is available, 1074 and new devices can be enrolled everywhere. Since all devices use 1075 the same trust anchor, a re-merge will be smooth. 1077 Merging two networks with different trust anchors requires the trust 1078 anchors to mutually trust each other (for example, by cross-signing). 1080 As long as the domain names are different, the addressing will not 1081 overlap (see Section 5.8). 1083 8. Self-Protection Properties 1085 As explained in Section 5, the ACP is based on secure channels built 1086 between devices that have mutually authenticated each other with 1087 their domain certificates. The channels themselves are protected 1088 using standard encryption technologies like DTLS or IPsec which 1089 provide additional authentication during channel establishment, data 1090 integrity and data confidentiality protection of data inside the ACP 1091 and in addition, provide replay protection. 1093 An attacker will therefore not be able to join the ACP unless having 1094 a valid domain certificate, also packet injection and sniffing 1095 traffic will not be possible due to the security provided by the 1096 encryption protocol. 1098 The remaining attack vector would be to attack the underlying AN 1099 protocols themselves, either via directed attacks or by denial-of- 1100 service attacks. However, as the ACP is built using link-local IPv6 1101 address, remote attacks are impossible. The ULA addresses are only 1102 reachable inside the ACP context, therefore unreachable from the data 1103 plane. Also, the ACP protocols should be implemented to be attack 1104 resistant and not consume unnecessary resources even while under 1105 attack. 1107 9. The Administrator View 1109 An ACP is self-forming, self-managing and self-protecting, therefore 1110 has minimal dependencies on the administrator of the network. 1111 Specifically, since it is independent of configuration, there is no 1112 scope for configuration errors on the ACP itself. The administrator 1113 may have the option to enable or disable the entire approach, but 1114 detailed configuration is not possible. This means that the ACP must 1115 not be reflected in the running configuration of devices, except a 1116 possible on/off switch. 1118 While configuration is not possible, an administrator must have full 1119 visibility of the ACP and all its parameters, to be able to do 1120 trouble-shooting. Therefore, an ACP must support all show and debug 1121 options, as for any other network function. Specifically, a network 1122 management system or controller must be able to discover the ACP, and 1123 monitor its health. This visibility of ACP operations must clearly 1124 be separated from visibility of data plane so automated systems will 1125 never have to deal with ACP aspect unless they explicitly desire to 1126 do so. 1128 Since an ACP is self-protecting, a device not supporting the ACP, or 1129 without a valid domain certificate cannot connect to it. This means 1130 that by default a traditional controller or network management system 1131 cannot connect to an ACP. See Section 6.1 for more details on how to 1132 connect an NMS host into the ACP. 1134 10. Security Considerations 1136 An ACP is self-protecting and there is no need to apply configuration 1137 to make it secure. Its security therefore does not depend on 1138 configuration. 1140 However, the security of the ACP depends on a number of other 1141 factors: 1143 o The usage of domain certificates depends on a valid supporting PKI 1144 infrastructure. If the chain of trust of this PKI infrastructure 1145 is compromised, the security of the ACP is also compromised. This 1146 is typically under the control of the network administrator. 1148 o Security can be compromised by implementation errors (bugs), as in 1149 all products. 1151 Fundamentally, security depends on correct operation, implementation 1152 and architecture. Autonomic approaches such as the ACP largely 1153 eliminate the dependency on correct operation; implementation and 1154 architectural mistakes are still possible, as in all networking 1155 technologies. 1157 11. IANA Considerations 1159 Section 5.5.3 describes ACP over dTLS, which requires a well-known 1160 UDP port. We request IANA to assign this UDP port for 'ACP over 1161 dTLS'. 1163 Section 5.5.4 describes an option for the channel negotiation, the 1164 'ACP channel type'. We request IANA to create a registry for 'ACP 1165 channel type'. 1167 The ACP channel type is a 8-bit unsigned integer. This document only 1168 assigns the first value. 1170 Number | Channel Type | RFC 1171 ---------+-----------------------------------+------------ 1172 0 | GRE tunnel protected with | this document 1173 | IPsec transport mode | 1174 1-255 | reserved for future channel types | 1176 Section 5.8.2 describes a 'type' field in the base addressing scheme. 1177 We request IANA to create a registry for the 'ACP addressing scheme 1178 type'. The initial value and definition will be determined in a 1179 later version of this document. 1181 12. Acknowledgements 1183 This work originated from an Autonomic Networking project at Cisco 1184 Systems, which started in early 2010. Many people contributed to 1185 this project and the idea of the Autonomic Control Plane, amongst 1186 which (in alphabetical order): Ignas Bagdonas, Parag Bhide, Balaji 1187 BL, Alex Clemm, Yves Hertoghs, Bruno Klauser, Max Pritikin, Ravi 1188 Kumar Vadapalli. 1190 Further input and suggestions were received from: Rene Struik, Brian 1191 Carpenter, Benoit Claise. 1193 13. Change log [RFC Editor: Please remove] 1195 13.1. Initial version 1197 First version of this document: draft-behringer-autonomic-control- 1198 plane 1200 13.2. draft-behringer-anima-autonomic-control-plane-00 1202 Initial version of the anima document; only minor edits. 1204 13.3. draft-behringer-anima-autonomic-control-plane-01 1206 o Clarified that the ACP should be based on, and support only IPv6. 1208 o Clarified in intro that ACP is for both, between devices, as well 1209 as for access from a central entity, such as an NMS. 1211 o Added a section on how to connect an NMS system. 1213 o Clarified the hop-by-hop crypto nature of the ACP. 1215 o Added several references to GDNP as a candidate protocol. 1217 o Added a discussion on network split and merge. Although, this 1218 should probably go into the certificate management story longer 1219 term. 1221 13.4. draft-behringer-anima-autonomic-control-plane-02 1223 Addresses (numerous) comments from Brian Carpenter. See mailing list 1224 for details. The most important changes are: 1226 o Introduced a new section "overview", to ease the understanding of 1227 the approach. 1229 o Merged the previous "problem statement" and "use case" sections 1230 into a mostly re-written "use cases" section, since they were 1231 overlapping. 1233 o Clarified the relationship with draft-ietf-anima-stable- 1234 connectivity 1236 13.5. draft-behringer-anima-autonomic-control-plane-03 1238 o Took out requirement for IPv6 --> that's in the reference doc. 1240 o Added requirement section. 1242 o Changed focus: more focus on autonomic functions, not only virtual 1243 out of band. This goes a bit throughout the document, starting 1244 with a changed abstract and intro. 1246 13.6. draft-ietf-anima-autonomic-control-plane-00 1248 No changes; re-submitted as WG document. 1250 13.7. draft-ietf-anima-autonomic-control-plane-01 1252 o Added some paragraphs in addressing section on "why IPv6 only", to 1253 reflect the discussion on the list. 1255 o Moved the data-plane ACP out of the main document, into an 1256 appendix. The focus is now the virtually separated ACP, since it 1257 has significant advantages, and isn't much harder to do. 1259 o Changed the self-creation algorithm: Part of the initial steps go 1260 into the reference document. This document now assumes an 1261 adjacency table, and domain certificate. How those get onto the 1262 device is outside scope for this document. 1264 o Created a new section 6 "workarounds for non-autonomic nodes", and 1265 put the previous controller section (5.9) into this new section. 1266 Now, section 5 is "autonomic only", and section 6 explains what to 1267 do with non-autonomic stuff. Much cleaner now. 1269 o Added an appendix explaining the choice of RPL as a routing 1270 protocol. 1272 o Formalised the creation process a bit more. Now, we create a 1273 "candidate peer list" from the adjacency table, and form the ACP 1274 with those candidates. Also it explains now better that policy 1275 (Intent) can influence the peer selection. (section 4 and 5) 1277 o Introduce a section for the capability negotiation protocol 1278 (section 7). This needs to be worked out in more detail. This 1279 will likely be based on GRASP. 1281 o Introduce a new parameter: ACP tunnel type. And defines it in the 1282 IANA considerations section. Suggest GRE protected with IPSec 1283 transport mode as the default tunnel type. 1285 o Updated links, lots of small edits. 1287 13.8. draft-ietf-anima-autonomic-control-plane-02 1289 o Added explicitly text for the ACP channel negotiation. 1291 o Merged draft-behringer-anima-autonomic-addressing-02 into this 1292 document, as suggested by WG chairs. 1294 13.9. draft-ietf-anima-autonomic-control-plane-03 1296 o Changed Neighbor discovery protocol from GRASP to mDNS. Bootstrap 1297 protocol team decided to go with mDNS to discover bootstrap proxy, 1298 and ACP should be consistent with this. Reasons to go with mDNS 1299 in bootstrap were a) Bootstrap should be reuseable also outside of 1300 full anima solutions and introduce as few as possible new 1301 elements. mDNS was considered well-known and very-likely even pre- 1302 existing in low-end devices (IoT). b) Using GRASP both for the 1303 insecure neighbor discovery and secure ACP operatations raises the 1304 risk of introducing security issues through implementation issues/ 1305 non-isolation between those two instances of GRASP. 1307 o Shortened the section on GRASP instances, because with mDNS being 1308 used for discovery, there is no insecure GRASP session any longer, 1309 simplifying the GRASP considerations. 1311 o Added certificate requirements for ANIMA in section 5.1.1, 1312 specifically how the ANIMA information is encoded in 1313 subjectAltName. 1315 o Deleted the appendix on "ACP without separation", as originally 1316 planned, and the paragraph in the main text referring to it. 1318 o Deleted one sub-addressing scheme, focusing on a single scheme 1319 now. 1321 o Included information on how ANIMA information must be encoded in 1322 the domain certificate in Section 5.1. 1324 o Editorial changes, updated draft references, etc. 1326 13.10. draft-ietf-anima-autonomic-control-plane-04 1328 Changed discovery of ACP neighbor back from mDNS to GRASP after 1329 revisiting the L2 problem. Described problem in discovery section 1330 itself to justify. Added text to explain how ACP discovery relates 1331 to BRSKY (bootstrap) discovery and pointed to Michael Richardsons 1332 draft detailing it. Removed appendix section that contained the 1333 original explanations why GRASP would be usedul (current text is 1334 meant to be better). 1336 14. References 1338 [I-D.ietf-anima-bootstrapping-keyinfra] 1339 Pritikin, M., Richardson, M., Behringer, M., Bjarnason, 1340 S., and K. Watsen, "Bootstrapping Remote Secure Key 1341 Infrastructures (BRSKI)", draft-ietf-anima-bootstrapping- 1342 keyinfra-04 (work in progress), October 2016. 1344 [I-D.ietf-anima-grasp] 1345 Bormann, C., Carpenter, B., and B. Liu, "A Generic 1346 Autonomic Signaling Protocol (GRASP)", draft-ietf-anima- 1347 grasp-08 (work in progress), October 2016. 1349 [I-D.ietf-anima-reference-model] 1350 Behringer, M., Carpenter, B., Eckert, T., Ciavaglia, L., 1351 Pierre, P., Liu, B., Nobre, J., and J. Strassner, "A 1352 Reference Model for Autonomic Networking", draft-ietf- 1353 anima-reference-model-02 (work in progress), July 2016. 1355 [I-D.ietf-anima-stable-connectivity] 1356 Eckert, T. and M. Behringer, "Using Autonomic Control 1357 Plane for Stable Connectivity of Network OAM", draft-ietf- 1358 anima-stable-connectivity-01 (work in progress), July 1359 2016. 1361 [I-D.richardson-anima-6join-discovery] 1362 Richardson, M., "GRASP discovery of Registrar by Join 1363 Assistant", draft-richardson-anima-6join-discovery-00 1364 (work in progress), October 2016. 1366 [RFC4122] Leach, P., Mealling, M., and R. Salz, "A Universally 1367 Unique IDentifier (UUID) URN Namespace", RFC 4122, 1368 DOI 10.17487/RFC4122, July 2005, 1369 . 1371 [RFC4193] Hinden, R. and B. Haberman, "Unique Local IPv6 Unicast 1372 Addresses", RFC 4193, DOI 10.17487/RFC4193, October 2005, 1373 . 1375 [RFC5082] Gill, V., Heasley, J., Meyer, D., Savola, P., Ed., and C. 1376 Pignataro, "The Generalized TTL Security Mechanism 1377 (GTSM)", RFC 5082, DOI 10.17487/RFC5082, October 2007, 1378 . 1380 [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S., 1381 Housley, R., and W. Polk, "Internet X.509 Public Key 1382 Infrastructure Certificate and Certificate Revocation List 1383 (CRL) Profile", RFC 5280, DOI 10.17487/RFC5280, May 2008, 1384 . 1386 [RFC5952] Kawamura, S. and M. Kawashima, "A Recommendation for IPv6 1387 Address Text Representation", RFC 5952, 1388 DOI 10.17487/RFC5952, August 2010, 1389 . 1391 [RFC6347] Rescorla, E. and N. Modadugu, "Datagram Transport Layer 1392 Security Version 1.2", RFC 6347, DOI 10.17487/RFC6347, 1393 January 2012, . 1395 [RFC6550] Winter, T., Ed., Thubert, P., Ed., Brandt, A., Hui, J., 1396 Kelsey, R., Levis, P., Pister, K., Struik, R., Vasseur, 1397 JP., and R. Alexander, "RPL: IPv6 Routing Protocol for 1398 Low-Power and Lossy Networks", RFC 6550, 1399 DOI 10.17487/RFC6550, March 2012, 1400 . 1402 [RFC6762] Cheshire, S. and M. Krochmal, "Multicast DNS", RFC 6762, 1403 DOI 10.17487/RFC6762, February 2013, 1404 . 1406 [RFC6763] Cheshire, S. and M. Krochmal, "DNS-Based Service 1407 Discovery", RFC 6763, DOI 10.17487/RFC6763, February 2013, 1408 . 1410 [RFC7404] Behringer, M. and E. Vyncke, "Using Only Link-Local 1411 Addressing inside an IPv6 Network", RFC 7404, 1412 DOI 10.17487/RFC7404, November 2014, 1413 . 1415 [RFC7575] Behringer, M., Pritikin, M., Bjarnason, S., Clemm, A., 1416 Carpenter, B., Jiang, S., and L. Ciavaglia, "Autonomic 1417 Networking: Definitions and Design Goals", RFC 7575, 1418 DOI 10.17487/RFC7575, June 2015, 1419 . 1421 [RFC7576] Jiang, S., Carpenter, B., and M. Behringer, "General Gap 1422 Analysis for Autonomic Networking", RFC 7576, 1423 DOI 10.17487/RFC7576, June 2015, 1424 . 1426 Appendix A. Background on the choice of routing protocol 1428 In a pre-standard implementation, the "IPv6 Routing Protocol for Low- 1429 Power and Lossy Networks (RPL, [RFC6550] was chosen. This 1430 Appendix explains the reasoning behind that decision. 1432 Requirements for routing in the ACP are: 1434 o Self-management: The ACP must build automatically, without human 1435 intervention. Therefore routing protocol must also work 1436 completely automatically. RPL is a simple, self-managing 1437 protocol, which does not require zones or areas; it is also self- 1438 configuring, since configuration is carried as part of the 1439 protocol (see Section 6.7.6 of [RFC6550]). 1441 o Scale: The ACP builds over an entire domain, which could be a 1442 large enterprise or service provider network. The routing 1443 protocol must therefore support domains of 100,000 nodes or more, 1444 ideally without the need for zoning or separation into areas. RPL 1445 has this scale property. This is based on extensive use of 1446 default routing. RPL also has other scalability improvements, 1447 such as selecting only a subset of peers instead of all possible 1448 ones, and trickle support for information synchronisation. 1450 o Low resource consumption: The ACP supports traditional network 1451 infrastructure, thus runs in addition to traditional protocols. 1452 The ACP, and specifically the routing protocol must have low 1453 resource consumption both in terms of memory and CPU requirements. 1454 Specifically, at edge nodes, where memory and CPU are scarce, 1455 consumption should be minimal. RPL builds a destination-oriented 1456 directed acyclic graph (DODAG), where the main resource 1457 consumption is at the root of the DODAG. The closer to the edge 1458 of the network, the less state needs to be maintained. This 1459 adapts nicely to the typical network design. Also, all changes 1460 below a common parent node are kept below that parent node. 1462 o Support for unstructured address space: In the Autonomic 1463 Networking Infrastructure, node addresses are identifiers, and may 1464 not be assigned in a topological way. Also, nodes may move 1465 topologically, without changing their address. Therefore, the 1466 routing protocol must support completely unstructured address 1467 space. RPL is specifically made for mobile ad-hoc networks, with 1468 no assumptions on topologically aligned addressing. 1470 o Modularity: To keep the initial implementation small, yet allow 1471 later for more complex methods, it is highly desirable that the 1472 routing protocol has a simple base functionality, but can import 1473 new functional modules if needed. RPL has this property with the 1474 concept of "objective function", which is a plugin to modify 1475 routing behaviour. 1477 o Extensibility: Since the Autonomic Networking Infrastructure is a 1478 new concept, it is likely that changes in the way of operation 1479 will happen over time. RPL allows for new objective functions to 1480 be introduced later, which allow changes to the way the routing 1481 protocol creates the DAGs. 1483 o Multi-topology support: It may become necessary in the future to 1484 support more than one DODAG for different purposes, using 1485 different objective functions. RPL allow for the creation of 1486 several parallel DODAGs, should this be required. This could be 1487 used to create different topologies to reach different roots. 1489 o No need for path optimisation: RPL does not necessarily compute 1490 the optimal path between any two nodes. However, the ACP does not 1491 require this today, since it carries mainly non-delay-sensitive 1492 feedback loops. It is possible that different optimisation 1493 schemes become necessary in the future, but RPL can be expanded 1494 (see point "Extensibility" above). 1496 Authors' Addresses 1498 Michael H. Behringer (editor) 1499 Cisco Systems 1500 Building D, 45 Allee des Ormes 1501 Mougins 06250 1502 France 1504 Email: mbehring@cisco.com 1506 Toerless Eckert 1508 Email: tte+ietf@cs.fau.de 1509 Steinthor Bjarnason 1510 Arbor Networks 1511 2727 South State Street, Suite 200 1512 Ann Arbor MI 48104 1513 United States 1515 Email: sbjarnason@arbor.net