| Internet-Draft | Problems and Requirements of Addressing | October 2021 |
| Li, et al. | Expires 28 April 2022 | [Page] |
This document presents a detailed analysis of the problems and requirements of network addressing in "Internet in space" for terrestrial users. It introduces the basics of satellite mega-constellations, terrestrial terminals/ground stations, and their inter-networking. Then it explicitly analyzes how space-terrestrial mobility yeilds challenges for the logical topology, addressing, and their impact on routing. The requirements of addressing in the space-terrestrial network are discussed in detail, including uniqueness, stability, locality, scalability, efficiency and backward compatibility with terrestrial Internet. The problems and requirements of network addressing in space-terrestrial networks are finally outlined.¶
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 28 April 2022.¶
Copyright (c) 2021 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 future Internet is up in the sky. We have seen a rocket-fast deployment of mega-constellations with 100s-10,000s of low-earth-orbit (LEO) satellites, such as Starlink [STARLINK], Kuiper [KUIPER] and OneWeb [ONEWEB]. These constellations promise competitive low latency and high capacity to terrestrial networks. They expand global high-speed Internet to remote areas that were not reachable by terrestrial networks, resulting in a tens-of-billions-of-dollar market with 3.7 billion users in rural areas[ITU-Measure], developing countries, aircraft, or oceans.¶
A salient feature for LEO mega-constellations is their high relative motions to the rotating earth. Unlike geosynchronous satellite or terrestrial networks, each LEO satellite moves fast (e.g., 28,080 km/h for Starlink), causing short-lived coverage for terrestrial users (less than 3 minutes). This yields diverse challenges for the traditional network designs.¶
This memo outlines the problems and requirements of addressing in integrated space-terrestrial network. It starts with the basics of satellite mega-constellations, terrestrial ground stations/terminals, and their inter-networking. It analyzes how high space-terrestrial relative motions yields challenges for logical topology, addressing and their impacts on routing. Then it discusses the requirements of network addressing in space-terrestrial network for uniqueness, stability, locality, scalability, efficiency and backward compatibility with terrestrial Internet.¶
GSO: Geosynchronous orbit (at the altitude of 35,786 km).¶
NGSO: Non-geosynchronous orbit.¶
LEO: Low Earth Orbit (at the altitude of 180-2,000 km).¶
MEO: Medium Earth Orbit (at the altitude of 180-35,786 km).¶
ISL: Inter Satellite Link.¶
NAT: Network Address Translation.¶
GS: Ground Station, a device on ground connecting the satellite.¶
FIB: Forwarding Information Base.¶
As shown in Figure 1, a space-terrestrial network for terrestrial users consists of the space segment and ground segment. The space segment includes the satellite or constellation. The ground segment comprises satellite terminals and ground stations.¶
Satellites can be classified based on their relative motions to the earth. A satellite can operate at the geosynchronous orbit (GSO, at about 35,786 km altitude) or non-geosynchronous orbits, such as low earth orbits (LEO, <=2,000km) and medium earth orbits (MEO, between 2,000 km and 35,786 km). Satellites at higher altitudes offer broader coverage, while satellites at lower altitudes move faster.¶
Historically, communications in space were dominated by GSO satellites. As shown in Table 1, GSO offers excellent coverage at high altitudes, but at the cost of long space-terrestrial RTT (>=200ms) and low bandwidth (<=10Mbps, due to bit errors in long distance transmission). Instead, recent efforts seek to adopt satellites at lower non-geosynchronous orbits, with a special interest in low-earth orbits. Together with Ku (12-18 GHz) and Ka (26.5-40GHz) bands, these satellites promise competitive bandwidth and latency to terrestrial networks( [LOWLATENCY-ROUTING-SPACE], [SPACE-RACE], [NETWORK-TOPO-DESIGN]). Due to low coverage for each LEO satellite, a mega-constellation is necessary to retain global coverage. Table 2 exemplifies popular LEO mega-costellations in operation. They are enabled by recent advances in satellite miniaturization and rocket reusability.¶
+-----------+ +-----------+
| Satellite |_______ISL_______| Satellite | (Space Segment)
_ _|___________| _ _|___________|
/| \ /|
/ | \ / |
/ (Bent pipe) \ | /
/ _ _\| /
+---------+ +------------+
| Mobile | | Ground |
| Station | | Station | (Ground Segment)
|_________| |____________|
| Orbit | Altitude (km) | RTT (ms) | |
|---|---|---|---|
| GSO | GEO | 35,786 | 240 |
| NGSO | MEO | 2,000-35,786 | 12-240 |
| LEO | 500-2,000 | 2-12 | |
| Constellation | Num. satellites | Num. orbits | Altitude (km) |
|---|---|---|---|
| Starlink | 1584 | 72 | 540/550 |
| 720 | 36 | 570 | |
| 348 | 6 | 560 | |
| 172 | 4 | 560 | |
| Kuiper | 1156 | 34 | 630 |
| 1296 | 36 | 610 | |
| 784 | 28 | 590 | |
| Telesat | 351 | 27 | 1015 |
| 1320 | 40 | 1325 | |
| Iridium | 66 | 6 | 780 |
Terrestrial users access satellite networks via terminals (e.g., satellite phones, onboard dishes, IoT endpoints) or ground stations. Ground stations can serve as network gateways (e.g., carrier-grade NAT in Starlink [STARLINK-CGNAT] and Kuiper [KUIPER-CGNAT]) and remote satellite controllers (e.g., telemetry, tracking, orbital update commands, or centralized routing control).¶
Early satellite communications favor the simple "bent-pipe-only" model (Figure 1), i.e., satellites only relay terrestrial users' radio signals to the fixed ground stations without ISLs or routing. This model has been popular in GSO satellites with broad coverage (2G GMR [GEO-MOBILE-RADIO-INTERFACE] [ETSI-TS-101], 3G BGAN [BGAN] [ETSI-TS-102], and DVB-S [SATELLITE-COMMUNICATIONS]), and recently adopted by LEO satellites in OneWeb (4G) [ONEWEB] and 5G NTN [STUDY-NR-SUPPORT] [SOLUTION-NR-NTN]. However, this model suffers from low LEO satellite coverage. To access the network, both terrestrial users and ground stations must reside inside the satellite's coverage. Due to each LEO satellite's low coverage, most users in remote areas with sparse or no ground stations cannot be served. As shown in Table 3, under current Starlink (no ISLs so far) and ground station deployments ([STARLINK-GS-FOUND], [AZURE-GS], [AWS-GS], [GOOGLE-DATA-CENTER], [AZURE-CLOUD-STARLINK], [STARLINK-GS-MAP]), 27%-52% global populations cannot be served by the "bent-pipe-only" model (depending on how many satellites each ground station can simultaneously associate to). Most under-served users are from remote areas (e.g., Africa), thus causing revenue loss for operators. Deploying dense ground stations in these remote areas could mitigate this. However, it is expensive and lowers commercial competitive advantages to terrestrial networks.¶
Instead, modern LEO mega-constellations favor satellite routing to expand global coverage or enable Internet backbones [Giuliari20Internet]. To date, inter-satellite links are still under early adoption([BEIDOU-TEST], [TheVerge-STARLINK-SPEED]). The recent "burn on re-entry" regulations from FCC also slows down the adoption of ISLs[SPACEX-CLAIM]. As a near-term remedy, ground station-assisted routing is currently adopted. There are two variants. The GS-as-gateway is adopted by Starlink and Kuiper. Each ground station is a carrier-grade NAT that offers private IP[RFC0791] for terrestrial users. The GS-as-relay [USE-GROUND-RELAY] mitigates ISLs with ground station-assisted routing, but is vulnerable to intermittent space-terrestrial links in Ku/Ka-bands. Like the "bent-pipe only" model (Figure 1), both heavily rely on ubiquitous ground station deployments in remote areas and even oceans, thus lowering competitive edges to terrestrial networks. Instead, we take a forward-looking view to exploring inter-satellite routing for its long-term success.¶
| Global | Africa | Oceania | South America | Asia | European | North America | |
|---|---|---|---|---|---|---|---|
| 1-SAT assicoation | 48.71% | 19.52% | 42.85% | 49.63% | 43.49% | 91.00% | 87.50% |
| 2-SAT assicoation | 57.30% | 24.37% | 56.58% | 53.90% | 55.91% | 94.33% | 91.23% |
| 4-SAT assicoation | 67.04% | 26.13% | 60.31% | 63.16% | 71.34% | 95.46% | 95.04% |
| 8-SAT assicoation | 73.04% | 29.17% | 60.68% | 65.65% | 80.28% | 96.91% | 98.86% |
In terrestrial and GEO satellite networks, the logical network topology, addresses, and routes are mostly stationary due to fixed infrastructure. Instead, LEO mega-constellations hardly enjoy this luxury, whose satellites move at high speeds (about 28,080 km/h). The earth's rotation further complicates the relative motions between space and ground. In this section, we will analyze how high relative motion between space and ground challenges addressing due to topology instability, and its impact on routing[INTERNET-IN-SPACE].¶
High physical mobility incurs frequent link churns between space and terrestrial nodes, thus causing frequent logical network topology changes. For all mega-constellations in Table 2, the topology changes every 10s of seconds[INTERNET-IN-SPACE]. The link churn populates with more satellites and ground stations.¶
In terrestrial mobile networks (e.g., 4G/5G), such physical link churn can be masked by handoffs without incurring logical topology changes. This method works based on two premises. First, all link churns occur at the last-hop radio due to user mobility, without affecting the infrastructure topology. Second, all cellular infrastructure nodes are fixed, resulting in a stable logical topology as "anchors".¶
However, neither premise holds in non-geosynchronous constellations. Instead, infrastructure mobility between satellites and ground stations becomes a norm rather than an exception. This voids cellular handoffs' merits to avoid propagation of physical link churns to logical network topology: They are designed for user mobility only, and heavily rely on the fixed infrastructure as "anchors." Therefore, 5G NTN lists satellite handoffs as an unsolved problem ([STUDY-NR-SUPPORT], [SOLUTION-NR-NTN]), and the latest 3GPP 5G release 17 defers its mobility support for satellites [TEC-SPECI-GROUP-MEETING] due to significant architectural changes. While Starlink uses handoffs to migrate physical links between satellites and ground stations (every 15s [STARLINK-CGNAT]), its logical topology and routing are still be repeatedly updated at high costs.¶
Each space/terrestrial node has two notions of "locations": The logical location in its topological address, and the physical location in reality. With repetitive topology changes, a static network address can hardly ensure its logical location in the topology is consistent with the fast-moving node's physical location in reality. Then to correctly forward data, a network should choose one of the following designs:¶
Dynamic address updates¶
A node can repetitively re-bind its physical location to its logical network address, thus incurring frequent address updates or re-binding. Under high mobility, this could severely disrupt user experiences or incur heavy signaling overhead. Table 4 and Table 5 project the address update frequency when using legacy IP addresses[RFC0791] for logical interfaces. In this scheme, the terrestrial users' logical IP[RFC0791] address changes if it re-associates to a new satellite (thus new interfaces and subnets) to retain its Internet access. Due to high LEO satellite mobility, each user is forced to change its logical IP address[RFC0791] every 133-510s. Every second, we observe 2,082-7,961 global users per second should change their IP addresses.¶
| Starlink | Telesat | Kuiper | Iridium |
|---|---|---|---|
| Every 133s | Every 510s | Every 510s | Every 458s |
| Starlink | Telesat | Kuiper | Iridium |
|---|---|---|---|
| 7961 | 2082 | 5673 | 2379 |
Static address binding to a fixed gateway¶
This is adopted by the cellular networks and Starlink [STARLINK-CGNAT] and Kuiper's[KUIPER-CGNAT] initial rollouts. Each user gets a static address from the remote ground station (via carrier-grade NAT), which masks the external address changes and redirects users' traffic. This mitigates user address updates, but cannot avoid gateway's external address updates when changing satellite interfaces (detailed below). It also incurs detours and long routing latencies for remote users from ground stations (e.g., 18,000 km detours and 370 ms extra delays in [lai2021icnp]).¶
The inconsistent locations in addressing further impact the network routing. As space and terrestrial infrastructure nodes physically move fast, the logical routing in cyberspace expires frequently. It must be updated frequently, thus threatening various routing schemes:¶
Distributed routing: Repetitive re-convergence.¶
In distributed routing, network nodes distribute topology information to others, locally compute forwarding tables, and eventually reach a global consensus on routing paths (i.e., convergence). Before global routing convergence, there is no guaranteed network reachability. With high mobility, each LEO satellite can only offer very short-lived access for a ground station(<=3 minutes in Starlink). Frequent topology updates cause repetitive routing re-convergence and thus low network usability. For intra-domain routing (e.g., OSPF[RFC2328], IS-IS[RFC1142]), most mega-constellations suffer from low network usability. Even the the size of constellation is samll, the network needs more than fifty seconds to converge after each handoff while using OSPF[TIMESLOT-DIVISION]. For inter-domain routing (e.g., BGP[RFC4271]), [Giuliari20Internet] and [NETWORK-IN-HEAVEN] show frequent logical topology changes cause BGP[RFC4271] re-peering, thus sharpening the instability of global Internet routing.¶
Centralized routing: Repetitive global updates.¶
In the centralized routing, a ground station predicts the temporal evolution of topology based on satellites' orbital patterns, divides it into a series of semi-static topology snapshots, schedules the forthcoming global routing tables for each snapshot, and remotely updates the routing tables to all satellites (e.g., via SDN[RFC7426], MPLS[RFC3031], or SRv6[RFC8754]). LEO mega-constellations pose stresses on centralized routing's scalability, stability, and complexity. Due to frequent link churns, the topology snapshots and FIBs explode with more satellites, ground stations, and faster mobility. Moreover, every satellite should locally load these new FIBs upon snapshot changes, which is vulnerable to transient global routing inconsistencies and thus black holes or loops.¶
Except from the basic properties like clusterability, network addressing in space-terrestrial network should also meet the following requirements:¶
In integrated space-terrestrial networks, each user's address should be globally unique. This property calls for address allocation and duplicate address detection mechanisms.¶
This property ensures that the addressing of each node does not change with the movement of users or satellites. With this property, location-based routing will be more stable, avoiding routing convergence caused by the high dynamics of integrated space-terrestrial network. It also reduces the frequency of network addresses updates and reduces the impact on users' network services.¶
For any two users or satellites, this property ensures that if their addresses are closer, their actual physical distances should be closer. It not only guarantees the unified logical and physical locations, but also simplifies the design and implementation of location-based routing.¶
The addressing of integrated space-terrestrial network should be designed to accommodate as many satellites and terrestrial nodes as possible. What's more, it should scale to more satellites and terrestrial users if needed.¶
In order to avoid frequent addresses updates, the design of addressing in space-terrestrial network should not require static address binding to remote gateways (e.g., carrier-grade NAT). The addressing method should ensure consistent cyber-physical locations, thus easing physically shortest paths without detours.¶
To ensure seamless expansion of the global Internet ecosystem, the addressing in space-terrestrial network should be backward compatible with terrestrial Internet. It should be compatible the standard IPv6 addressing formats, and facilitates inter-networking to external networks without modifying terrestrial infrastructure. For backward compatibility with IPv4, we recommend adopting 4over6 transition for integrated space-terrestrial networks.¶
This memo includes no request to IANA.¶
The present memo does not introduce any new technology and/or mechanism and as such does not introduce any security threat to the TCP/IP protocol suite.¶