| Network Working Group | F. Templin, Ed. |
| Internet-Draft | G. Saccone |
| Intended status: Informational | Boeing Research & Technology |
| Expires: July 15, 2019 | G. Dawra |
| A. Lindem | |
| V. Moreno | |
| Cisco Systems, Inc. | |
| January 11, 2019 |
A Simple BGP-based Mobile Routing System for the Aeronautical Telecommunications Network
draft-ietf-rtgwg-atn-bgp-01.txt
The International Civil Aviation Organization (ICAO) is investigating mobile routing solutions for a worldwide Aeronautical Telecommunications Network with Internet Protocol Services (ATN/IPS). The ATN/IPS will eventually replace existing communication services with an IPv6-based service supporting pervasive Air Traffic Management (ATM) for Air Traffic Controllers (ATC), Airline Operations Controllers (AOC), and all commercial aircraft worldwide. This informational document describes a simple and extensible mobile routing service based on industry-standard BGP to address the ATN/IPS requirements.
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."
This Internet-Draft will expire on July 15, 2019.
Copyright (c) 2019 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
The worldwide Air Traffic Management (ATM) system today uses a service known as Aeronautical Telecommunications Network based on Open Systems Interconnection (ATN/OSI). The service is used to augment controller to pilot voice communications with rudimentary short text command and control messages. The service has seen successful deployment in a limited set of worldwide ATM domains.
The International Civil Aviation Organization [ICAO] is now undertaking the development of a next-generation replacement for ATN/OSI known as Aeronautical Telecommunications Network with Internet Protocol Services (ATN/IPS). ATN/IPS will eventually provide an IPv6-based [RFC8200] service supporting pervasive ATM for Air Traffic Controllers (ATC), Airline Operations Controllers (AOC), and all commercial aircraft worldwide. As part of the ATN/IPS undertaking, a new mobile routing service will be needed. This document presents an approach based on the Border Gateway Protocol (BGP) [RFC4271].
Aircraft communicate via wireless aviation data links that typically support much lower data rates than terrestrial wireless and wired-line communications. For example, some Very High Frequency (VHF)-based data links only support data rates on the order of 32Kbps and an emerging L-Band data link that is expected to play a key role in future aeronautical communications only supports rates on the order of 1Mbps. Although satellite data links can provide much higher data rates during optimal conditions, like any other aviation data link they are subject to errors, delay, disruption, signal intermittence, degradation due to atmospheric conditions, etc. The well-connected ground domain ATN/IPS network should therefore treat each safety-of-flight critical packet produced by (or destined to) an aircraft as a precious commodity and strive for an optimized service that provides the highest possible degree of reliability.
The ATN/IPS is an IPv6-based overlay network that assumes a worldwide connected Internetworking underlay for carrying tunneled ATM communications. The Internetworking underlay could be manifested as a private collection of long-haul backbone links (e.g., fiber optics, copper, SATCOM, etc.) interconnected by high-performance networking gear such as bridges, switches, and routers. Such a private network would need to connect all ATN/IPS participants worldwide, and could therefore present a considerable cost for a large-scale deployment of new infrastructure. Alternatively, the ATN/IPS could be deployed as a secured overlay over the existing global public Internet. For example, ATN/IPS nodes could be deployed as part of an SD-WAN or an MPLS-WAN that rides over the public Internet via secured tunnels. Further details of the Internetworking underlay design are out of scope for this document.
The ATN/IPS further assumes that each aircraft will receive an IPv6 Mobile Network Prefix (MNP) that accompanies the aircraft wherever it travels. ICAO is further proposing to assign each aircraft an entire /56 MNP for numbering its on-board networks. ATCs and AOCs will likewise receive IPv6 prefixes, but they would typically appear in static (not mobile) deployments such as air traffic control towers, airline headquarters, etc. Throughout the rest of this document, we therefore use the term "MNP" when discussing an IPv6 prefix that is delegated to any ATN/IPS end system, including ATCs, AOCs, and aircraft. We also use the term Mobility Service Prefix (MSP) to refer to an aggregated prefix assigned to the ATN/IPS by an Internet assigned numbers authority, and from which all MNPs are delegated (e.g., up to 2^32 IPv6 /56 MNPs could be delegated from an IPv6 /24 MSP).
Connexion By Boeing [CBB] was an early aviation mobile routing service based on dynamic updates in the global public Internet BGP routing system. Practical experience with the approach has shown that frequent injections and withdrawals of MNPs in the Internet routing system can result in excessive BGP update messaging, slow routing table convergence times, and extended outages when no route is available. This is due to both conservative default BGP protocol timing parameters (see Section 6) and the complex peering interconnections of BGP routers within the global Internet infrastructure. The situation is further exacerbated by frequent aircraft mobility events that each result in BGP updates that must be propagated to all BGP routers in the Internet that carry a full routing table.
We therefore consider an approach using a BGP overlay network routing system where a private BGP routing protocol instance is maintained between ATN/IPS Autonomous System (AS) Border Routers (ASBRs). The private BGP instance does not interact with the native BGP routing system in the connected Internetworking underlay, and BGP updates are unidirectional from "stub" ASBRs (s-ASBRs) to a small set of "core" ASBRs (c-ASBRs) in a hub-and-spokes topology. No extensions to the BGP protocol are necessary.
The s-ASBRs for each stub AS connect to a small number of c-ASBRs via dedicated high speed links and/or tunnels across the Internetworking underlay using industry-standard encapsulations (e.g., Generic Routing Encapsulation (GRE) [RFC2784], IPsec [RFC4301], etc.). In particular, tunneling must be used when neighboring ASBRs are separated by multiple Internetworking underlay hops.
The s-ASBRs engage in external BGP (eBGP) peerings with their respective c-ASBRs, and only maintain routing table entries for the MNPs currently active within the stub AS. The s-ASBRs send BGP updates for MNP injections or withdrawals to c-ASBRs but do not receive any BGP updates from c-ASBRs. Instead, the s-ASBRs maintain default routes with their c-ASBRs as the next hop, and therefore hold only partial topology information.
The c-ASBRs connect to other c-ASBRs using internal BGP (iBGP) peerings over which they collaboratively maintain a full routing table for all active MNPs currently in service. Therefore, only the c-ASBRs maintain a full BGP routing table and never send any BGP updates to s-ASBRs. This simple routing model therefore greatly reduces the number of BGP updates that need to be synchronized among peers, and the number is reduced further still when intradomain routing changes within stub ASes are processed within the AS instead of being propagated to the core. BGP Route Reflectors (RRs) [RFC4456] can also be used to support increased scaling properties.
The remainder of this document discusses the proposed BGP-based ATN/IPS mobile routing service.
The terms Autonomous System (AS) and Autonomous System Border Router (ASBR) are the same as defined in [RFC4271].
The following terms are defined for the purposes of this document:
The proposed ATN/IPS routing system comprises a private BGP instance coordinated in an overlay network via tunnels between neighboring ASBRs over the Internetworking underlay. The overlay does not interact with the native BGP routing system in the connected underlying Internetwork, and each c-ASBR advertises only a small and unchanging set of MSPs into the Internetworking underlay routing system instead of the full dynamically changing set of MNPs. (For example, when the Internetworking underlay is the global public Internet the c-ASBRs advertise the MSPs in the public BGP Internet routing system.)
In a reference deployment, one or more s-ASBRs connect each stub AS to the overlay using a shared stub AS Number (ASN). Each s-ASBR further uses eBGP to peer with one or more c-ASBRs. All c-ASBRs are members of the same core AS, and use a shared core ASN. Globally-unique public ASNs could be assigned, e.g., either according to the standard 16-bit ASN format or the 32-bit ASN scheme defined in [RFC6793].
The c-ASBRs use iBGP to maintain a synchronized consistent view of all active MNPs currently in service. Figure 1 below represents the reference deployment. (Note that the figure shows details for only two s-ASBRs (s-ASBR1 and s-ASBR2) due to space constraints, but the other s-ASBRs should be understood to have similar Stub AS, MNP and eBGP peering arrangements.) The solution described in this document is flexible enough to extend to these topologies.
........................................................... . . . (:::)-. <- Stub ASes -> (:::)-. . . MNPs-> .-(:::::::::) .-(:::::::::) <-MNPs . . `-(::::)-' `-(::::)-' . . +-------+ +-------+ . . |s-ASBR1+-----+ +-----+s-ASBR2| . . +--+----+ eBGP \ / eBGP +-----+-+ . . \ \/ / . . \eBGP / \ /eBGP . . \ / \ / . . +-------+ +-------+ . . eBGP+-----+c-ASBR |...|c-ASBR +-----+eBGP . . +-------+ / +--+----+ +-----+-+ \ +-------+ . . |s-ASBR +/ iBGP\ (:::)-. /iBGP \+s-ASBR | . . +-------+ .-(::::::::) +-------+ . . . .-(::::::::::::::)-. . . . . (:::: Core AS :::) . . . +-------+ `-(:::::::::::::)-' +-------+ . . |s-ASBR +\ iBGP/`-(:::::::-'\iBGP /+s-ASBR | . . +-------+ \ +-+-----+ +----+--+ / +-------+ . . eBGP+-----+c-ASBR |...|c-ASBR +-----+eBGP . . +-------+ +-------+ . . / \ . . /eBGP \eBGP . . / \ . . +---+---+ +-----+-+ . . |s-ASBR | |s-ASBR | . . +-------+ +-------+ . . . . . . <------------ Internetworking Underlay --------------> . ............................................................
Figure 1: Reference Deployment
Each c-ASBR configures a black-hole route for each of its MSPs. By black-holing the MSPs, the c-ASBR will maintain forwarding table entries only for the MNPs that are currently active, and packets destined to all other MNPs will correctly incur ICMPv6 Destination Unreachable messages [RFC4443] due to the black hole route. (This is the same behavior as for ordinary BGP routers in the Internet when they receive packets for which there is no route available.) The c-ASBRs do not send eBGP updates for MNPs to s-ASBRs, but instead originate a default route. In this way, s-ASBRs have only partial topology knowledge (i.e., they know only about the active MNPs currently within their stub ASes) and they forward all other packets to c-ASBRs which have full topology knowledge.
Scaling properties of this ATN/IPS routing system are limited by the number of BGP routes that can be carried by the c-ASBRs. A 2015 study showed that BGP routers in the global public Internet at that time carried more than 500K routes with linear growth and no signs of router resource exhaustion [BGP]. A more recent network emulation study also showed that a single c-ASBR can accommodate at least 1M dynamically changing BGP routes even on a lightweight virtual machine. Commercially-available high-performance dedicated router hardware can support many millions of routes.
Therefore, assuming each c-ASBR can carry 1M or more routes, this means that at least 1M ATN/IPS end system MNPs can be serviced by a single set of c-ASBRs and that number could be further increased by using RRs and/or more powerful routers. Another means of increasing scale would be to assign a different set of c-ASBRs for each set of MSPs. In that case, each s-ASBR still peers with one or more c-ASBRs from each set of c-ASBRs, but the s-ASBR institutes route filters so that it only sends BGP updates to the specific set of c-ASBRs that aggregate the MSP. In this way, each set of c-ASBRs maintains separate routing and forwarding tables so that scaling is distributed across multiple c-ASBR sets instead of concentrated in a single c-ASBR set. For example, a first c-ASBR set could aggregate an MSP segment A::/32, a second set could aggregate B::/32, a third could aggregate C::/32, etc. The union of all MSP segments would then constitute the collective MSP(s) for the entire ATN/IPS.
In this way, each set of c-ASBRs services a specific set of MSPs that they inject into the Internetworking underlay native routing system, and each s-ASBR configures MSP-specific routes that list the correct set of c-ASBRs as next hops. This design also allows for natural incremental deployment, and can support initial medium-scale deployments followed by dynamic deployment of additional ATN/IPS infrastructure elements without disturbing the already-deployed base. For example, a few more c-ASBRs could be added if the MNP service demand ever outgrows the initial deployment. For larger-scale applications (such as unmanned air vehicles and terrestrial vehicles) even larger scales can be accommodated by adding more c-ASBRs.
Radio Access Networks (RANs) connect end system Clients such as aircraft, ATCs, AOCs etc. to the ATN/IPS routing system. Clients may connect to multiple RANs at once, for example, when they have both satellite and cellular data links activated simultaneously. Clients may further move between RANs in a manner that is perceived as a network layer mobility event. Clients could therefore employ a multilink/mobility routing service such as those discussed in Section 7.
Clients register all of their active data link connections with their serving s-ASBRs as discussed in Section 3. Clients may connect to s-ASBRs either directly, or via a Proxy at the edge of the RAN.
Figure 2 shows the ATN/IPS RAN model where Clients connect to RANs via aviation data links. Clients register their RAN addresses with a nearby s-ASBR, where the registration process may be brokered by a Proxy at the edge of the RAN.
Data Link "A" +--------+ Data Link "B"
+----------- | Client |-----------+
/ +--------+ \
/ \
/ \
(:::)-. (:::)-.
.-(:::::::::) <- Radio Access Networks -> .-(:::::::::)
`-(::::)-' `-(::::)-'
+-------+ +-------+
... | Proxy | ............................ | Proxy | ...
. +-------+ +-------+ .
. ^^ ^^ .
. || || .
. || +--------+ || .
. ++============> | s-ASBR | <============++ .
. +--------+ .
. | eBGP .
. (:::)-. .
. .-(::::::::) .
. .-(::: ATN/IPS :::)-. .
. (::::: BGP Routing ::::) .
. `-(:: System ::::)-' .
. `-(:::::::-' .
. .
. .
. <------------- Internetworking Underlay -------------> .
............................................................
Figure 2: ATN/IPS RAN Architecture
When a Client logs into a RAN, it specifies a nearby s-ASBR that it has selected to connect to the ATN/IPS. (Selection of a nearby s-ASBR could be through consulting a geographically-keyed static host file, through a DNS lookup, etc.) The login process is transparently brokered by a Proxy at the border of the RAN, which then conveys the connection request to the s-ASBR via tunneling across the Internetworking underlay. The s-ASBR then registers the address of the Proxy as the address for the Client, and the Proxy forwards the s-ASBR's reply to the Client. If the Client connects to multiple RANs, the s-ASBR will register the addresses of all Proxies as addresses through which the Client can be reached.
The s-ASBR represents all of its active Clients as MNP routes in the ATN/IPS BGP routing system. The s-ASBR's stub AS therefore consists of the set of all of its active Clients (i.e., the stub AS is a logical construct and not a physical construct). The s-ASBR injects the MNPs of its active Clients and withdraws the MNPs of its departed Clients via BGP updates to c-ASBRs. Since Clients are expected to remain associated with their current s-ASBR for extended periods, the level of MNP injections and withdrawals in the BGP routing system will be on the order of the numbers of network joins, leaves and s-ASBR handovers for aircraft operations (see: Section 6). It is important to observe that fine-grained events such as Client mobility and Quality of Service (QoS) signaling are coordinated only by Proxies and the Client's current s-ASBRs, and do not involve other ASBRs in the routing system. In this way, intradomain routing changes within the stub AS are not propagated into the rest of the ATN/IPS BGP routing system.
ATN/IPS end systems will frequently need to communicate with correspondents associated with other s-ASBRs. In the BGP peering topology discussed in Section 3, this can initially only be accommodated by including multiple tunnel segments in the forwarding path. In many cases, it would be desirable to eliminate extraneous tunnel segments from this "dogleg" route so that packets can traverse a minimum number of tunneling hops across the Internetworking underlay. ATN/IPS end systems could therefore employ a route optimization service according to the mobility service employed (see: Section 7).
A route optimization example is shown in Figure 3 and Figure 4 below. In the first figure, multiple tunneled segments between Proxys and ASBRs are necessary to convey packets between Clients associated with different s-ASBRs. In the second figure, the optimized route tunnels packets directly between Proxys without involving the ASBRs.
+---------+ +---------+
| Client1 | | Client2 |
+---v-----+ +-----^---+
* *
* *
(:::)-. (:::)-.
.-(:::::::::) <- Radio Access Networks -> .-(:::::::::)
`-(::::)-' `-(::::)-'
+--------+ +--------+
... | Proxy1 | .......................... | Proxy2 | ...
. +--------+ +--------+ .
. ** ** .
. ** ** .
. ** ** .
. +---------+ +---------+ .
. | s-ASBR1 | | s-ASBR2 | .
. +--+------+ +-----+---+ .
. \ ** Dogleg ** / .
. eBGP\ ** Route ** /eBGP .
. \ **==============** / .
. +---------+ +---------+ .
. | c-ASBR1 | | c-ASBR2 | .
. +---+-----+ +----+----+ .
. +--------------+ .
. iBGP .
. .
. <------------- Internetworking Underlay -------------> .
............................................................
Figure 3: Dogleg Route Before Optimization
+---------+ +---------+
| Client1 | | Client2 |
+---v-----+ +-----^---+
* *
* *
(:::)-. (:::)-.
.-(:::::::::) <- Radio Access Networks -> .-(:::::::::)
`-(::::)-' `-(::::)-'
+--------+ +--------+
... | Proxy1 | .......................... | Proxy2 | ...
. +------v-+ +--^-----+ .
. * * .
. *================================* .
. .
. +---------+ +---------+ .
. | s-ASBR1 | | s-ASBR2 | .
. +--+------+ +-----+---+ .
. \ / .
. eBGP\ /eBGP .
. \ / .
. +---------+ +---------+ .
. | c-ASBR1 | | c-ASBR2 | .
. +---+-----+ +----+----+ .
. +--------------+ .
. iBGP .
. .
. <------------- Internetworking Underlay -------------> .
............................................................
Figure 4: Optimized Route
The number of eBGP peering sessions that each c-ASBR must service is proportional to the number of s-ASBRs in the system. Network emulations with lightweight virtual machines have shown that a single c-ASBR can service at least 100 eBGP peerings from s-ASBRs that each advertise 10K MNP routes (i.e., 1M total). It is expected that robust c-ASBRs can service many more peerings than this - possibly by multiple orders of magnitude. But even assuming a conservative limit, the number of s-ASBRs could be increased by also increasing the number of c-ASBRs. Since c-ASBRs also peer with each other using iBGP, however, larger-scale c-ASBR deployments may need to employ an adjunct facility such as BGP Route Reflectors (RRs)[RFC4456].
The number of aircraft in operation at a given time worldwide is likely to be significantly less than 1M, but we will assume this number for a worst-case analysis. Assuming a worst-case average 1 hour flight profile from gate-to-gate with 10 service region transitions per flight, the entire system will need to service at most 10M BGP updates per hour (2778 updates per second). This number is within the realm of the peak BGP update messaging seen in the global public Internet today [BGP2]. Assuming a BGP update message size of 100 bytes (800bits), the total amount of BGP control message traffic to a single c-ASBR will be less than 2.5Mbps which is a nominal rate for modern data links.
Industry standard BGP routers provide configurable parameters with conservative default values. For example, the default hold time is 90 seconds, the default keepalive time is 1/3 of the hold time, and the default MinRouteAdvertisementinterval is 30 seconds for eBGP peers and 5 seconds for iBGP peers (see Section 10 of [RFC4271]). For the simple mobile routing system described herein, these parameters can and should be set to more aggressive values to support faster neighbor/link failure detection and faster routing protocol convergence times. For example, a hold time of 3 seconds and a MinRouteAdvertisementinterval of 0 seconds for both iBGP and eBGP.
Each c-ASBR will be using eBGP both in the ATN/IPS and the Internetworking Underlay with the ATN/IPS unicast IPv6 routes resolving over Internetworking Underlay routes. Consequently, c-ASBRs and potentially s-ASBRs will need to support separate local ASes for the two BGP routing domains and routing policy or assure routes are not propagated between the two BGP routing domains. From a conceptual and operational standpoint, the implementation should provide isolation between the two BGP routing domains (e.g., separate BGP instances).
Stub ASes maintain intradomain routing information for mobile node clients, and are responsible for all localized mobility signaling without disturbing the BGP routing system. Clients can enlist the services of a candidate mobility service such as Mobile IPv6 (MIPv6) [RFC6275], LISP [I-D.ietf-lisp-rfc6830bis] and AERO [I-D.templin-intarea-6706bis] according to the service offered by the stub AS. Further details of mobile routing services are out of scope for this document.
The BGP routing topology described in this document has been modeled in realistic network emulations showing that at least 1 million MNPs can be propagated to each c-ASBR even on lightweight virtual machines. No BGP routing protocol extensions need to be adopted.
This document does not introduce any IANA considerations.
ATN/IPS ASBRs on the open Internet are susceptible to the same attack profiles as for any Internet nodes. For this reason, ASBRs should employ physical security and/or IP securing mechanisms such as IPsec [RFC4301], TLS [RFC5246], etc.
ATN/IPS ASBRs present targets for Distributed Denial of Service (DDoS) attacks. This concern is no different than for any node on the open Internet, where attackers could send spoofed packets to the node at high data rates. This can be mitigated by connecting ATN/IPS ASBRs over dedicated links with no connections to the Internet and/or when ASBR connections to the Internet are only permitted through well-managed firewalls.
ATN/IPS s-ASBRs should institute rate limits to protect low data rate aviation data links from receiving DDoS packet floods.
This document does not include any new specific requirements for mitigation of DDoS.
This work is aligned with the FAA as per the SE2025 contract number DTFAWA-15-D-00030.
This work is aligned with the NASA Safe Autonomous Systems Operation (SASO) program under NASA contract number NNA16BD84C.
This work is aligned with the Boeing Information Technology (BIT) MobileNet program.
The following individuals contributed insights that have improved the document: Erik Kline, Tony Li, Alexandre Petrescu, Pascal Thubert.
| [RFC4271] | Rekhter, Y., Li, T. and S. Hares, "A Border Gateway Protocol 4 (BGP-4)", RFC 4271, DOI 10.17487/RFC4271, January 2006. |
| [RFC4443] | Conta, A., Deering, S. and M. Gupta, "Internet Control Message Protocol (ICMPv6) for the Internet Protocol Version 6 (IPv6) Specification", STD 89, RFC 4443, DOI 10.17487/RFC4443, March 2006. |
| [RFC4456] | Bates, T., Chen, E. and R. Chandra, "BGP Route Reflection: An Alternative to Full Mesh Internal BGP (IBGP)", RFC 4456, DOI 10.17487/RFC4456, April 2006. |
| [RFC8200] | Deering, S. and R. Hinden, "Internet Protocol, Version 6 (IPv6) Specification", STD 86, RFC 8200, DOI 10.17487/RFC8200, July 2017. |
<< RFC Editor - remove prior to publication >>
Changes from -00 to -01:
Status as of 08/30/2018: