Zeqi Lai, Hewu Li, Qian Wu, Qiang Ni, Mingyang Lv, Jihao Li, Jianping Wu, Jun Liu, Yuanjie Li
©SHUTTERSTOCK.COM/DUKESN
Futuristic 6G technologies will integrate emerging low-Earth orbit (LEO) megaconstellations into terrestrial networks, promising to provide ubiquitous, low-latency and high-throughput network services on-demand. However, several unique characteristics of satellites (e.g., high dynamics and error-prone operational environments) make it very challenging to unleash the potential of megaconstellations and accomplish these aforementioned promises.
In this article, we study a new computation paradigm, called space edge computing (SEC), which wisely combines the advanced computation, network, and storage resources on emerging satellites to achieve pervasive and performant on-demand services for futuristic 6G communications. Specifically, we investigate SEC in three steps. First, we explore the technical feasibility of SEC, and envision a novel integrated satellite-cloud architecture that collaboratively exploits “space edges” and “terrestrial clouds” to provide on-demand services anytime, anywhere. Second, through a series of quantitative experiments, we showcase SEC’s potential on enhancing the network availability and performance for several quintessential applications. Finally, we conclude with the practical challenges facing the SEC, and accordingly outline a list of corresponding new directions for future research.
Emerging megaconstellation consisting of thousands of LEO satellites is one of the key building blocks for futuristic 6G networks. Recently, we have witnessed a renaissance in the space industry, which significantly reduces the cost of launching spacecrafts to space and stimulates an exponential increase in constructing Internet megaconstellations. As compared to their predecessor, recent LEO satellites can be equipped with high-resolution sensors, space-grade multicore processors, high-speed communication links, and multifunctional space software. Satellites in free space accomplish wide-area network coverage and improve the network availability for terrestrial users, especially for those users in remote or rural regions. While there are several existing technologies like citizens broadband radio service that can provide communication services in a flexible and low-cost manner without having to acquire spectrum licenses, these terrestrial communication techniques offer limited coverage and data rate as compared to upcoming broadband satellites.
To unlock the power of emerging broadband satellites, the concrete approach for internetworking LEO megaconstellations and terrestrial networks plays a critical role in the service quality of the entire integrated network. Figure 1 plots several existing approaches proposed for space-ground integration. First, current operational megaconstellations like SpaceX’s Starlink follow a “bent-pipe” architecture to integrate LEO satellites into ground facilities. Data from the ground are first transmitted to the satellites, which then send the data right back down again like a bent pipe. Second, another internetworking approach has been proposed in [5], in which ground stations work as access points or gateways for satellite networks [ground station access for satellite networks (GSAS)]. LEO satellites leverage intersatellite links (ISLs) to build space routes for long-haul communications. Data packets from users are first forwarded to an access ground station via terrestrial networks, and then forwarded to the satellite backbone in space. Finally, recent studies [7] have proposed a paradigm where users leverage satellite terminals to directly access the ISL-enabled satellite networks [satellite networks directly accessed by terrestrial users (SDAT)]. Satellites work not only as routers, but also as access points or gateways for ground users.
Figure 1 Existing design space of space–ground integration. GSAS: ground station access for satellite networks; SDAT: satellite networks directly accessed by terrestrial users.
However, due to a series of unique characteristics of emerging megaconstellations, it is still difficult for existing space-ground architectures to simultaneously accomplish pervasive, reliable, and low-latency network services. First, the bent-pipe architecture only uses satellites as relays to exchange data between users and ground stations, and its network accessibility is limited as it can only serve users close to certain ground stations. Second, GSAS is also accessibility-limited, as it requires well-deployed terrestrial networks between users and ground stations and cannot offer services for users in regions where terrestrial facilities are constrained. Finally, although SDAT can achieve pervasive network accessibility since it does not require a large number of geo-distributed ground stations, the user-satellite connectivity suffers from frequent changes due to LEO dynamics. Users’ IP addresses have to be frequently updated, making it difficult to guarantee stable and reliable routes and end-to-end transport connections.
In this article, we study a new computation paradigm, called SEC, and explore the potential of integrating SEC with terrestrial networks to accomplish pervasive, reliable, and low-latency service for futuristic 6G networks. The key idea behind SEC is to wisely combine and schedule advanced resources (e.g., computation, storage, and network resources) in emerging LEO satellites to satisfy various application requirements on demand.
We carry out our quest in three steps. First, we conduct an in-depth analysis on the feasibility of realizing SEC in emerging satellite systems and propose a new satellite–cloud integrated architecture in futuristic 6G paradigm which collaboratively exploits the heterogeneous computation, storage, and network resources in space edges and terrestrial cloud data centers to enable on-demand service anytime, anywhere (see the “On-Demand Pervasive Service Above the Atmosphere: SEC” section).
Second, through a series of quantitative experiments driven by public details of real constellation design and orbital information, we quantitatively show the potential opportunities enabled by SEC, and demonstrate how quintessential applications can benefit from the collaborative architecture, regarding network accessibility, reliability and performance (see the “New Opportunities” section).
Finally, we highlight a collection of practical challenges facing the deployment and application of SEC, which are jointly involved by the high dynamics of LEO satellites and the failure-prone, resource-constrained space environment. We also conclude a list of corresponding future research problems (see the “Challenges Ahead and Future Directions” section). Future efforts are expected to cope with these challenges and improve the robustness and efficiency of SEC.
We first analyze several critical aspects about the feasibility of building edge-like service platforms in space.
Intersatellite and ground-satellite communications: The total capacity of satellite communication systems has increased significantly over the past decade. For example, during the beta test of Starlink, end users can achieve data speeds varying from 50 Mbps to 150 Mbps and latency from 20 ms to 40 ms in most available locations. Many constellations under construction also suggest the use of laser ISLs, which can potentially support up to tens or even hundreds of Gb/s data transmission rate for intersatellite communication [8]. Moreover, leading cloud providers such as Amazon and Microsoft are actively deploying their ground-stations-as-a-service platforms [4], allowing satellite operators to use ground services on a flexible pay-as-you-go basis with affordable costs, and without the need to deploy their own ground infrastructures. For instance, the reserved wide-band services provided by Amazon’s ground station services, which enable satellite operators to download their satellite data or build backup downlinks, are billed at a rate of US$15 maintenance cost per minute [1].
Onboard computation systems: Recently BAE Systems and Boeing are developing their next-generation high-performance multicore processor for space-computing platforms [6], promising to support complex computational workloads. Researchers also start to explore the feasibility of using commercial off-the-shelf (COTS) devices in space environments. In 2021, Hewlett Packard Enterprise announced the COTS Spaceborne Computer-2 [3], which costs less than US${\$}$20,000, to introduce edge computing and AI capabilities to international space stations (ISSs), focusing on astronaut health care and image processing, and so on. Similarly, European Space Agent (ESA) has launched two COTS Astro Pi units to ISSs for education-focused experiments in space. Each unit is based on a Raspberry Pi computer, which costs less than US$100. The latest progress mentioned previously demonstrates the feasibility of bringing performant, affordable, and compact computation systems to the outer space.
Storage, weight, and volume: In addition to the evolved computation and communication capabilities, future satellites may also provide data storage services in space. The storage capacity of spaceborne solid state recorder (SSR) has raised from 2 Gb in 2000 to the present 20 Tb per spacecraft. Taking Starlink as an example, the weight of a typical SSR is about 6–20 kg, which is about 2.3–7.6% of the weight of the latest launched Starlink satellite. The volume of SSR is about 250 mm × 250 mm × 250 mm, less than 3% of a Starlink satellite.
Launch capability: SpaceX has already been able to position 60 260-kg satellites into orbit in one launch. The current Falcon-9 first-stage booster is said to be capable of at least 10 flights, which further decreases the cost. The enhanced launch capability and reduced cost thus make it probably feasible to deploy megaconstellations with thousands of satellites in orbit within a few years and accomplish pervasive satellite coverage.
Thus, we envision a new integrated architecture for future wireless communication: integrating SEC with terrestrial networks for future 6G pervasive on-demand services. As shown in Figure 2, collectively the integrated architecture consists of three primary segments.
Figure 2 Integration of SEC and terrestrial networks for pervasive on-demand services at a global scale.
Global content distribution: Content distribution networks (CDNs) play an important role in today’s terrestrial networks. However, a recent measurement study [13] has revealed that the effectiveness of existing cloud-/edge-based CDNs is limited due to their insufficient infrastructure deployment and the unstable last-mile access. From a global perspective, there are still a large number of users suffering from poor content accessibility or high latency due to the scattered data center deployment.
SEC is promising to extend the availability and network performance of today’s terrestrial CDNs. Intuitively, contents can be stored in satellite caches, which are operated in LEO and close to terrestrial users. We conduct an experiment to quantitatively analyze the reachability and access latency from different ground locations to their closest LEO satellite in the first phase of Starlink constellation, which plans to launch about 4400 satellites in total. We quantify the reachability of a ground user by the number of reachable satellites in the current location of the user and quantify the access latency by the round-trip time (RTT) between a ground user and the satellite it connects to. According to the Federal Communications Commission filings, the angle at which a Starlink satellite is reachable is about 25°. Satellites in each shell interconnect with each other following a +Grid topology, in which each satellite connects two adjacent neighbors in the same orbit, and other two adjacent neighbors in the left/right orbit. Table 1 summarizes the primary constellation parameters, including the number of orbits, the number of satellites in each orbit, orbital height, and inclination. We calculate the latency by the method proposed in a recent satellite analysis tool [10].
Table 1 Primary constellation parameters.
Figure 3 shows the number of reachable satellites, and Figure 4 plots the RTT estimations in various locations that differed in latitudes. We observe that in most terrestrial locations, the number of simultaneously reachable satellites varies from 15 to 40. Since most satellites in the constellation are operated in the inclined orbits, the reachability concentrates on latitude range [60°S, 60°N]. In addition, since the orbital height is less than 600 km, all LEO constellations enable about 3–8-ms propagation RTT inside their coverage. Our experiment results indicate that if satellites are properly designed to cache Internet contents, allowing users to fetch data objects directly from the closest reachable satellite instead of fetching data from the remote original server, the content access latency can (potentially) be significantly reduced, especially for those users in remote or rural areas suffering from limited terrestrial networks. Note that it is also very important to protect data privacy and prevent leakage for satellite caches, as satellites are operated in a public, intermittent, and resource-constrained environment. We leave the data privacy issue in satellite edges as our future work.
Figure 3 The number of reachable satellites in different geo-distributed locations.
Figure 4 RTT (ms) estimations between Starlink satellites and ground users in different geo-distributed locations. H: orbit height. I: orbit inclination. N: number of satellites. The dotted/solid line represents the maximum/minimum values.
Wide-area real-time communication (RTC): In recent years, we have seen a dramatic rise in RTC applications, e.g., videoconferencing, interactive immersive applications (e.g., augmented reality/virtual reality), emerging remote surge, autonomous vehicles, and so on, with significantly stringent latency requirements on underlying networks. As the end-to-end latency is very critical for these applications, a well-known approach to optimize RTC latency is leveraging cloud-based RTC relays to forward RTC traffic [9]. This method leverages geographically distributed cloud data centers that are connected by an overlay network to construct highly performant wide-area networks.
SEC extends the optional set of RTC relays: Servers in either terrestrial cloud data centers or LEO satellites can be selected as RTC relays to control and forward RTC traffic. The potential benefit of using in-orbit relays is twofold. First, for long-distance communications (e.g., cross-continent RTC sessions), interconnected satellites can build obstacle-free, low-latency paths and avoid meandering routes. Second, laser links can communicate at the speed of light, which is about 47% faster as compared to that in existing terrestrial fibers.
To quantitatively demonstrate the potential latency reduction made by SEC for long-distance communication, we conduct an analysis to calculate and compare the achievable latency by terrestrial networks and by the satellite-cloud integrated architecture. Figure 5 plots the low-latency communication path for wide-area cross-continent sessions enabled by SEC. Specifically, dash arrows indicate the terrestrial Internet paths connecting two populated areas on different continents, while solid arrows refer to the path constructed by networked satellite edges. We identify these paths by the traceroute tool between two vantage points in corresponding cities. In practice, the terrestrial path is generated by Internet routing protocols [e.g., Border Gateway Protocol (BGP)], following specific policies in different autonomous systems. Solid arrows indicate the shortest paths built upon a large number of satellite edges. Here satellites are interconnected following the topology of Starlink’s first shell. Quantitatively, we find that SEC can achieve significantly reduced latency by avoiding meandering paths. In each case shown in Figure 5(a) and (b) respectively, the end-to-end propagation latency is about 106.05/124.8 ms over terrestrial paths and 34.6/47.8 ms over SEC paths. Since Starlink is still under heavy development today and we have very limited access to them, in this experiment we mainly focus on the propagation latency which is determined by the network topology and the speed of light.
Figure 5 Potential low-latency paths enabled by SEC. Dashed arrows indicate the default terrestrial path generated by today’s Internet routing protocol (e.g., BGP), and solid arrows refer to the low-latency paths built upon networked space edges. (a) Communication path from South Africa to Brazil. (b) Communication path from China to South Africa.
In addition to the pervasive and low-latency capability, recent technologies also enable high-throughput LEO communication for terrestrial users. For example, a recent measurement study [12] demonstrates that the current form of Starlink can provide 100–250 Mbps download speed, which is sufficient for high-definition video streaming. Combining the low-latency and high-throughput potential, SEC promises to support various emerging applications with delay- and bandwidth-sensitive requirements, such as self-driving, mixed reality, remote surgery, and so on, in the future. Due to the page limit, we leave the analysis on SEC’s application in other scenarios to our future work.
Typically, network reliability captures how long the network infrastructure is correctly functional without interruptions. In futuristic 6G networks which integrate LEO satellites into terrestrial networks, sustaining high reliability is still very difficult in the intermittent, error-prone environment.
Reliability strengthening agent: To eliminate frequent user address updates, SEC allows satellite operators to deploy reliability strengthening agent in satellites. In particular, for each terrestrial user, the agent works as an anchor which allocates and maintains addresses for terrestrial users in the dynamic environment to guarantee stable network connections. Figure 6 shows an example illustrating the basic idea of SEC-based agent. The network topology changes in each slot due to the LEO dynamics. Two ground users Src and Dst establish a communication session over the satellite network. One satellite is selected as the reliability strengthening agent, and it allocates unchanged addresses to Src and Dst. During the session, Src forwards packets to the agent first, and then forwards to the Dst. Although the ground-satellite connectivity changes from slot 1 to slot 3, the source-to-agent and agent-to-destination connections can be kept stable since terrestrial users do not change their addresses under the fluctuating topology.
Figure 6 Reliability strengthening agent based on space edges.
Network-driven fast recovery: In conventional networks, loss recovery is typically controlled by the sender. For example, a Transmission Control Protocol (TCP) sender estimates current loss rate based on the received TCP’s acknowledgements (ACKs) number and resends lost packets to guarantee reliable transmission. However, for long-haul communications in satellite networks where loss could be common and frequent, a sender-controlled recovery mechanism can be slow and inefficient. SEC can improve the recovery efficiency by adopting network-driven fast recovery. In particular, SEC can cache packets on error-prone paths. As shown in Figure 7, when a packet loss occurs, the sender-controlled mechanism needs to resend the packet from the sender, while the network-driven solution achieves fast and efficient packet recovery by resending the lost packet from an edge close to the receiver.
Figure 7 Network-driven loss recovery based on space edges.
SEC can also be used to improve the transmission efficiency for space missions, such as high-quality Earth observation (EO) services. In recent years, EO data are growing in size and variety at an exceptionally fast rate. Data collected by sensors in space needs to be downloaded to terrestrial control and operating centers for further data processing. Existing methods for delivering data from space can typically be classified into two categories: 1) leveraging ground stations to download data when a sensing satellite arrives at the transmission range of a certain ground station [15], or 2) exploiting geostationary satellites to build a two-hop bent-pipe path for data delivery.
EO services are used in many scenarios, including situations where EO data collected in orbit is expected to be delivered to the ground as soon as possible. Examples include disaster management, emergency response, remote surveillance and security. However, the aforementioned methods for EO data download suffer from limited transmission efficiency. In the former method 1) because sensing satellites move at high velocity during their orbit, a certain ground-satellite link can last for only a few minutes (e.g., 3 min or less) in one pass. It may take multiple passes (e.g., several hours or even days) to complete the entire EO data delivery. The latter method 2) exploits geostationary satellites to form a stable delivery path to forward EO data to the ground. While geostationary satellites working at 36,000 km altitude can relay and forward data for sensing satellites and facilitate long-duration stable communication, this method still suffers from high transmission completion time (TCT) due to the limited data rate from the sensing satellite to geostationary (GEO) satellite relays [14]. SEC enables two opportunities to improve the transmission efficiency for EO tasks.
Fast data delivery via networked space edges: We conduct a quantitative experiment based on real constellation information and EO data trace to demonstrate the effectiveness of SEC on reducing the TCT. We build a simulator which simulates sensing satellites based on the public information of the Dove EO constellation, which is currently one of the largest EO constellations, and simulates networked satellite edges based on the Starlink broadband constellation. Satellite edges interconnect with each other following the +Grid topology. The ground station topology is configured based on the information of real ground station infrastructures [2]. In our simulator, once a sensing satellite collects EO data, it connects to a nearby satellite edge, which then exploits the shortest path to forward EO data back to the ground data center.
Figure 8 plots the CDF of TCT of delivering 12-h EO data collected from Amazon Forest by networked SEC, and by other two existing methods, i.e., download by distributed ground station networks with different numbers of available ground stations (denoted as D-GSes), and download by GEO relays (denoted as GEO-Relay). In particular, TCT is calculated as the duration from the time when an EO satellite starts to download its data, to the time when the collected data are fully delivered to the ground. For ground station networks, we change the number of available ground stations, and the TCT is about 82–7161 s (2868.8 on average) under 50 ground stations, and 81–5546 s (2205.3 on average) even though 173 distributed ground stations are all used. This is because in the worst case it may take tens of minutes to hours for the sensing satellite to move into the transmission range, and the duration of the visible window of an LEO satellite in one pass is very limited. GEO relays offer stable data transmission, but the download latency is about 215–461 s (324.3 on average), limited by the insufficient forwarding data rate. SEC achieves 80–172 s (121.6 on average) download latency, and reduces up to 94.49% and 62.50% TCT on average as compared to the existing methods.
Figure 8 TCT of 12 h EO data of Amazon Forest from Dove EO constellation to terrestrial data centers.
Reducing bandwidth overhead by on-board intelligence: Onboard processing is another promising direction to improve the transmission efficiency of EO data delivery and other distant tasks. Emerging spaceborne computer systems facilitate the run of applications with high computational complexity such as data compression and deep learning [11]. For some remote missions, naively transferring all large-volume space data can impose significant bandwidth overhead on intersatellite and ground-satellite links and involve high communication power consumption. Exploiting onboard data compression or intelligence technologies to compress raw data, or detect and discard unnecessary information, can significantly reduce the burden on data transmission. For example, a sensing satellite observing the Earth can run object detection and classification algorithms to identify and discard unavailable images (e.g., cloudy pictures where targets on the ground are covered by clouds), and accordingly save the bandwidth overhead.
Unlike traditional mobile networks where only end users are mobile (e.g., cellular networks) or dynamics are limited to a local area (e.g., vehicular ad hoc networks), the high velocity of LEO satellites results in high dynamics in the core network infrastructure at a global scale. Such dynamics result in issues such as frequent address updates and connectivity disruptions, and accordingly impose significant challenges on sustaining stable and long-duration SEC services. We highlight two future research directions related to taming the dynamics in SEC.
Resilience techniques in failure-prone space environments: The space backbone network of the SEC architecture is exposed in the complex outer space environment, suffering from risks such as debris collisions and radiation hazards. All these factors can result in node or link failures in space. Therefore, future research is expected to study the resilience techniques for SEC, e.g., resilient network protocols and resource scheduling techniques, to accomplish robust and highly reliable SEC services in failure-prone environments.
Content distributions upon the satellite-cloud integrated architecture: Since SEC promises pervasive and low-latency content access, it is thus very important to decide how to collaboratively place Internet content (e.g., web content, video clips) on satellites and clouds. Essentially, satellite caches and cloud caches complement each other, in terms of their coverage, performance and cost. For example, satellite caches achieve better coverage especially for users in remote and rural regions, but also involve higher delivery cost as compared to cloud platforms. A wise content provision approach is expected to balance the cost and effectiveness for various content distribution tasks under the integrated architecture.
Space resources are still precious and limited as compared to well-provisioned and well-optimized terrestrial infrastructures. Thus, it is very important to properly and efficiently exploit the precious resource in space.
Space-ground collaborative computation: As we have introduced previously, in the SEC architecture, Earth-observing satellites can preprocess data acquired in space, e.g., through deep-learning models, to detect and discard unnecessary information. In particular, machine learning models contain a sequence of layers to sequentially process input images and extract features. However, a high-accuracy model for object detection or classification can still involve high computation overhead on power-limited satellite edges. To accelerate the EO data processing, the entire model can be properly divided into two parts collaboratively executed by satellites and terrestrial data centers respectively. Future works are expected to find the optimal division for various onboard processing tasks.
Typically, LEO satellites can spend over 30% of their orbital period under the Earth’s eclipse, during which they have to be powered by rechargeable batteries. Satellite systems typically include a number of subsystems working for various dedicated functions. The communication subsystem may take a large portion of the entire power budget. Although the batteries can be recharged by solar energy, the depth of discharge (DoD) can significantly affect the battery lifetime. Key technologies for SEC, including space routing, transmission control, resource scheduling and data processing are expected to be energy-efficient and battery-friendly to extend the lifetime of satellites.
Recent evolution in both spaceborne hardware and software envisions a new paradigm of computing service, i.e., SEC. This article explores SEC in three aspects. First, we analyze the feasibility of SEC and propose a novel integrated satellite-cloud architecture. Second, through a series of quantitative experiments, we identify several use cases demonstrating that SEC can potentially enhance the network accessibility, reliability, and performance for various applications. We finally conclude key technical challenges facing the SEC, and highlight related future directions.
This work was supported by the National Key R&D Program of China (2022YFB3105202), National Natural Science Foundation of China (NSFC 62132004), and Tsinghua University-China Telecom Joint Research Institute for Next Generation Internet Technology. The authors would like to thank all anonymous reviewers for their feedback which greatly improved the article.
Zeqi Lai (zeqilai@tsinghua.edu.cn) is an assistant professor at the Institute for Network Sciences and Cyberspace, Tsinghua University, Beijing 100084, China. Before joining Tsinghua University, he was a senior researcher at Tencent Media Lab from 2018 to 2019. He received his Ph.D. degree in computer science from Tsinghua University in 2018. His research interests include next-generation Internet architecture and protocols, integrated space and terrestrial networks, wireless and mobile computing, network security, and video streaming.
Hewu Li (lihewu@cernet.edu.cn) is an associate professor and assistant to the Dean of the Institute for Network Sciences and Cyberspace, Tsinghua University, Beijing 100084, China. He received his M.S. and Ph.D. degrees in computer science from Tsinghua University in 2001 and 2004, respectively. He His research interests include integrated space and terrestrial networks, mobile wireless network architecture, broadband wireless access technology, and mobility architecture in next-generation Internet.
Qian Wu (wuqian@cernet.edu.cn) is an associate professor in the Institute for Network Sciences and Cyberspace, Tsinghua University, Beijing 100084, China. She received her M.S. and Ph.D. degrees in computer science from Tsinghua University, in 2002 and 2006, respectively. Her research interests include the next-generation Internet architecture and protocols, integrated space-terrestrial networks, mobile and wireless networks, multipath transfer, and mobile multicast.
Qiang Ni (q.ni@lancaster.ac.uk) is a professor at the School of Computing and Communications, Lancaster University, LA1 4WA Lancaster, U.K. His research areas include future generation communications and networking, including green communications/networking, millimeter-wave wireless, cognitive radio systems, 5G/6G, SDN, cloud networks, edge computing, dispersed computing, Internet of Things, cyber physical systems, artificial intelligence/machine learning, and vehicular networks. He has authored or coauthored 300+ papers in these areas.
Mingyang Lv (lvmy20@mails.tsinghua.edu.cn) is currently working toward an M.S. degree at the institute for Network Sciences and Cyberspace, Tsinghua University, Beijing 100084, China. He received his B.S. degree in network engineering from Sun Yat-Sen University in 2018. His research interests mainly include big data distribution and routing in integrated space and terrestrial networks.
Jihao Li (lijh19@mails.tsinghua.edu.cn) is pursuing his Ph.D. degree in the Department of Computer Science and Technology, Tsinghua University, Beijing 100084, China. His current research areas include the architecture and routing of integrated space and terrestrial networks.
Jianping Wu (jianping@cernet.edu.cn) is the dean of the Institute for Network Sciences and Cyberspace and professor in the Department of Computer Science, Tsinghua University, Beijing 100084, China. He is also the director of CERNET National Network Center and CERNET Technical Board and the director of the National Engineering Laboratory for Next Generation Internet. He is a member of Chinese Academy of Engineering and International Fellow of the Royal Academy of Engineering of U.K.
Jun Liu (juneliu@tsinghua.edu.cn) is currently an assistant professor with the Institute of Network Sciences and Cyberspace, Tsinghua University, Beijing 100084, China, and a researcher with the Beijing National Research Center for Information Science and Technology. She was a postdoctoral research fellow from 2016–2018 in the CECA Department at Peking University. She received her Ph.D. degree from Northeastern University, Shenyang, China, in 2011. Her research interests focus on wireless networks and space-air-terrestrial networks.
Yuanjie Li (yuanjiel@tsinghua.edu.cn) is an assistant professor at the Institute for Network Sciences and Cyberspace, Tsinghua University, Beijing 100084, China. Previously, he was a researcher at Hewlett Packard Labs from 2018 to 2020. He received his Ph.D. degree in computer science from the University of California Los Angeles in 2017 and his B.E. degree in electronic engineering from Tsinghua University in 2012. His research interests include network systems and security, with a recent focus on satellite and mobile networking.
[1] “Aws ground station pricing.” Amazon. [Online] . Available: https://aws.amazon.com/ground-station/pricing/
[2] Satnogs Network. [Online] . Available: https://network.satnogs.org/
[3] D. Werner, “Spaceborne computer-2 completes 24 experiments on ISS,” Space News, Apr. 2022. [Online] . Available: https://spacenews.com/spaceborne-computer-2-symposium-2022/
[4] “Aws ground station as a service.” Amazon. [Online] . Available: https://aws.amazon.com/ground-station/
[5] G. Giuliari, T. Klenze, M. Legner, D. Basin, A. Perrig, and A. Singla, “Internet backbones in space,” ACM SIGCOMM Comput. Commun. Rev., vol. 50, no. 1, pp. 25–37, Jan. 2020, doi: 10.1145/3390251.3390256.
[6] E. W. Gretok, E. T. Kain, and A. D. George, “Comparative benchmarking analysis of next-generation space processors,” in Proc. IEEE Aerosp. Conf., 2019, pp. 1–16, doi: 10.1109/AERO.2019.8741914.
[7] M. Handley, “Delay is not an option: Low latency routing in space,” in Proc. 17th ACM Workshop Hot Topics Netw. (HotNets), 2018, pp. 85–91, doi: 10.1145/3286062.3286075.
[8] Y. Hauri, D. Bhattacherjee, M. Grossmann, and A. Singla, “‘Internet from space’ without inter-satellite links,” in Proc. 19th ACM Workshop Hot Topics Netw. (HotNets), 2020, pp. 205–211, doi: 10.1145/3422604.3425938.
[9] J. Jiang et al., “Via: Improving internet telephony call quality using predictive relay selection,” in Proc. ACM SIGCOMM Conf., 2016, pp. 286–299, doi: 10.1145/2934872.2934907.
[10] S. Kassing, D. Bhattacherjee, A. Baptista Águas, J. Eirik Saethre, and A. Singla, “Exploring the ‘Internet from space’ with Hypatia,” in Proc. ACM Internet Meas. Conf. (IMC), 2020, pp. 214–229, doi: 10.1145/3419394.3423635.
[11] V. Kothari, E. Liberis, and N. D. Lane, “The final frontier: Deep learning in space,” in Proc. 21st Int. Workshop Mobile Comput. Syst. Appl. (HotMobile), 2020, pp. 45–49, doi: 10.1145/3376897.3377864.
[12] F. Michel, M. Trevisan, D. Giordano, and O. Bonaventure, “A first look at starlink performance,” in Proc. ACM Internet Meas. Conf. (IMC), 2022, vol. 10, pp. 130–136, doi: 10.1145/3517745.3561416.
[13] N. Mohan, L. Corneo, A. Zavodovski, S. Bayhan, W. Wong, and J. Kangasharju, “Pruning edge research with latency shears,” in Proc. 19th ACM Workshop Hot Topics Netw. (HotNets), 2020, pp. 182–189, doi: 10.1145/3422604.3425943.
[14] T. Sobchak, D. W. Shinners, and H. Shaw, “NASA space network project operations management: Past, present and future for the tracking and data relay satellite constellation,” in Proc. SpaceOps Conf., 2018, pp. 1–12, doi: 10.2514/6.2018-2358.
[15] D. Vasisht and R. Chandra, “A distributed and hybrid ground station network for low earth orbit satellites,” in Proc. 19th ACM Workshop Hot Topics Netw. (HotNets), 2020, pp. 190–196, doi: 10.1145/3422604.3425926.
Digital Object Identifier 10.1109/MVT.2022.3221391