Space Microdatacenters¶
Earth observation (EO) has been a key task for satellites since the first time a satellite was put into space. The temporal and spatial resolution at which EO satellites take pictures has been increasing to support space-based applications, but this increases the amount of data each satellite generates. We observe that future EO satellites will generate so much data that this data cannot be transmitted to Earth due to the limited capacity of communication that exists between space and Earth. We show that conventional data reduction techniques such as compression [130] and early discard [54] do not solve this problem, nor does a direct enhancement of today’s RFbased infrastructure [136, 153] for space-Earth communication. We explore an unorthodox solution instead - moving to space the computation that would have happened on the ground. This alleviates the need for data transfer to Earth. We analyze ten non-longitudinal RGB and hyperspectral image processing Earth observation applications for their computation and power requirements and discover that these requirements cannot be met by the small satellites that dominate today’s EO missions. We make a case for space microdatacenters - large computational satellites whose primary task is to support in-space computation of EO data. We show that one 4KW space microdatacenter can support the computation need of a majority of applications, especially when used in conjunction with early discard. We do find, however, that communication between EO satellites and space microdatacenters becomes a bottleneck. We propose three space microdatacenter-communication co-design strategies – 𝑘 − 𝑙𝑖𝑠𝑡-based network topology, microdatacenter splitting, and moving space microdatacenters to geostationary orbit that alleviate the bottlenecks and enable effective usage of space microdatacenters.
Introduction¶
The ability to launch satellites into space and then control them to accomplish a wide variety of tasks such as navigation [56], communication [64], forecasting [119], early warning [111], reconnaissance [107], broadcasting [106], scientific research [46], signals intelligence [109, 154], weapons delivery [73], and Earth observation [127] has been one of the most wondrous achievements of humankind. These satellites have different volumes (0.01 m 3 to 916 m 3 ) and weights (1.26 kg to 420 000 kg) and are placed into outer space at different altitudes above the Earth (274 km to 35 786 km) in different orbits (low Earth orbit [47], geostationary orbit [142], sunsynchronous orbit (SSO) [34], etc.) using launch vehicles [38, 50]. These satellites have different sources of power generation (none - for passive satellites [126], solar panels [123], radioisotopic thermoelectric generators [122], etc.) to support their functionality, use transponders [55] for communication to Earth-based ground stations [95], and work either alone or together as a group (often called a constellation [147]).
Earth observation (EO) has been a key task for satellites since inception. EO satellites image the Earth using camera [127], radar [67], lidar [116], photometer [140], or atmospheric instruments [36] in order to support a variety of scientific [12], military [109, 154], and commercial [55] applications. As imaging satellites, they are often placed in low Earth orbit for high data resolution (though some EO satellites are placed in a geostationary orbit [142] for uninterrupted coverage or in a SSO for consistent lighting during imaging [11]), and transmit their images to Earth-based ground stations for further processing. Following Sputnik-1 [117], the first satellite ever launched, thousands of EO satellites have been placed in space to support different applications [102]. A vast number of future satellite launches are also devoted to Earth observation [94] to support a fast growing Earth observation industry [94].
A key parameter for an EO satellite is the resolution at which it takes its pictures. Increasingly Earth observation space missions are being planned with aggressive goals of spatial and temporal resolution (Section 3) to support emerging EO applications such as forest fire detection [148], realtime video [134], conflict zone monitoring [28], tasking [40], warning systems for early responders [156], and tracking of events such as Earthquakes [159], hurricanes [51], and tornadoes [43], as well as objects such as aircraft [77] and missiles [31]. Even traditional EO applications such as flood monitoring [155], traffic monitoring [86], mapping [44], etc., seek higher resolutions requirements now. Mapping a narrow path in a dense urban area easily requires sub-meter resolution [158], for example. Fig. 2 shows how spatial resolution of EO satellites has improved over the decades.
In this paper, we observe that the amount of data that future high resolution Earth observation satellites will generate will be so massive that data cannot simply be transmitted to the Earth considering present or projected ground station capacity (Section 3). The limited number of ground stations on the Earth limit the total amount of data that can be transmitted. At current costs, the monetary cost of transmission will also be prohibitive (Section 3).
We first evaluate two techniques (Section 4) that have been previously proposed to reduce the amount of data transmitted to the Earth - compression [130] and early discard [54] - to address the problem (Fig. 1b). We show that compression or early discard may not provide sufficient data reduction for many high resolution space missions either alone or in conjunction. We also consider (Section 4) if today’s RF-based communication infrastructure can be enhanced to support high resolution space missions. We show that practical RF-based satellite antennas may not support the needs for many such missions. The number of channels needed to be supported on the ground may also be unrealistic.
We explore an unorthodox solution instead (Section 5) - whenever possible, move the computation that would have happened on the ground to space itself. If we are able to perform the computation in space itself, only insights, not raw sensor data, may need to be transmitted to the ground alleviating the need for massive data transfer to the ground for high resolution applications.
We analyze ten emerging non-longitudinal RGB and hyperspectral image processing Earth observation applications that process high resolution satellite data. We estimate for these applications their computation and power requirements at different resolutions. We find that small satellites which dominate Earth observation today, cannot support many of these applications, especially at high resolutions, as these satellites cannot generate enough power to support the power requirements of these applications. While early discard helps reduce the power requirements, the reduction is not enough to support many of these applications.
With the above in mind, we make a case for space microdatacenters (SµDCs) for high resolution Earth observation space missions (Section 6). A SµDC (Fig. 1c) is a relatively large computational satellite whose primary task is to support in-space computation on data generated by the observation satellites. The power generation capability for the SµDC is commensurate with the amount of computation supported by the SµDC. Inter-satellite links (ISLs) are used to offload the data generated by the observation satellites to the SµDC.
We consider the SµDC requirements for a 64-satellite constellation of Earth observation satellites for 4KW SµDCs based on NVIDIA RTX 3090-class processors. We show (Section 6) that one 4 kW SµDC can support the computation needs for a majority of our applications for most resolutions, especially when used in conjunction with early discard.
We do find, however, that communication between the observation satellites and the SµDCs becomes a bottleneck (Section 7). We propose three SµDC-communication co-design strategies – 𝑘 −𝑙𝑖𝑠𝑡based network topology, SµDC splitting, and moving SµDCs to geostationary orbit – to alleviate this bottleneck and effectively use these SµDCs (Section 8). Finally, we analyze the impact of placement and chip architecture on SµDC design and performance.
This paper makes the following contributions:
• We show that future high resolution Earth observation missions will generate so much data that the generated data cannot be transmitted to the Earth considering present or projected ground station capacity or considering the transmission costs.
• We show that compression, early discard, or antenna scaling have limited effectiveness at addressing the problem.
• We explore moving the Earth-based computation that computes on EO data into space and show that this computation cannot be performed on the typically small EO satellites since these satellites cannot meet the corresponding power requirements.
• We make a quantitative case for SµDCs that are designed to run the Earth-based computation in space. We show that a 4 kW SµDC can support a majority of the applications if communication bottlenecks can be alleviated.
• We present multiple SµDC-communication co-design strategies (new connection topologies, SµDC splitting, moving SµDCs to geostationary orbit) that alleviate the communication bottlenecks of SµDCs.