Commercial data centers are increasingly exploring space, and hyperscalers are offering their services on a service model. The world is planning to increase the satellite constellation and, in the near future, to use it to start providing IoT services.
Computing in orbit
Mankind more and more actively uses outer space, the number of satellites is approaching 5 thousand. They are used for navigation and communication, financial transactions, observations of space objects, weather and surface of the Earth. Users of space services control forests and fields, track the movements of vessels and help in rescue operations. The volume of data collected by satellites is growing rapidly. Data centers are required for data processing and it is advisable to place such data centers directly in space.
The technology of peripheral computing (Edge Computing), which has already been tested in hard-to-reach areas of the Earth, looks promising. Space edge data centers will process primary data, while data centers on the ground will analyze aggregated information and provide management. On-site pre-processing will increase the speed of decision-making and reduce the load on communication channels with the cloud in the head data center.
Main advantages of space data centers
- Increase of information processing efficiency up to real-time processing.
- Reducing the load on expensive and low-bandwidth communication channels with the Earth.
Learn also: SaaS marketing
Space from the cloud
It is convenient for customers to receive space services from the cloud of a familiar hyperscaler, and work is already underway in this direction: cloud providers are establishing cooperation with satellite launch companies. Capella Space’s collaboration with AWS, which began in 2020, has resulted in AWS Aerospace & Satellite Solutions’ AWS Ground Station-based satellite management cloud service (Figure 1, 2). Using the cloud service, customers can process data without having to worry about setting up their own ground station and managing its infrastructure. And the AWS open data catalog includes image sets from Sentinel-2, Landsat and GOES satellite constellations.
Figure 1. AWS Ground Station provides customer interaction with satellites
A partnership with Google Cloud is helping Ilon Musk’s company SpaceX process data from low-orbit Starlink satellites. SpaceX is actively working on inter-satellite link technology, which could eventually lead to separate satellites to perform in-orbit calculations and build an orbital distributed computing network.
Source: Capella Space
Figure 2. Capella Space satellite
Satellite operator SES announced that Microsoft will become the first cloud provider of the O3b mPOWER constellation operating in medium Earth orbit. As part of the global Azure Space initiative, Microsoft is working with SpaceX, satellite operator SES, KSAT, Viasat and US EIectrodynamics to develop solutions that link space and ground facilities, including data centers. Using the MS Azure cloud allows storing and analyzing vast amounts of data for commercial satellite orbit control and space debris monitoring.
In September 2022, SES and Microsoft announced an expanded partnership in a satellite virtualization program that includes all nodes of the ecosystem, from software-defined radios and terminals to virtualized networks and peripheral computing.
In the first satellites, software was created for specialized server hardware of space objects. The use of standard server hardware allows the use of conventional commercial software, including virtualization tools and cloud platforms, to move to software-defined hyperconverged systems that enable different companies to share orbital computing resources and software repurpose satellites for their tasks. In particular, they can create edge-clouds for multiple users to share a satellite, such as processing satellite imagery data or monitoring distributed sensors for Internet of Things applications.
Startups that are building satellite micro data centers have already emerged. In 2019, California-based Vector announced plans to launch the GSky-1 software-defined satellite. The goal of the project, in collaboration with the University of Southern California (USC), is to deploy the GalacticSky cloud platform on microsatellites and provide users with space computing power on a service model. The lightweight Vector launch vehicle was to orbit a constellation of interconnected microsatellites, which were placed in sturdy containers of micro data centers (up to 16 virtual machines in each). Such a microsatellite-container may tumble, lose communication with the Earth, break down, but the performance of a single container data center will not affect the operation of the system as a whole. Based on a cluster of microsatellites with the same orbit, the GalacticSky cloud platform was supposed to be deployed.
The launch planned for the end of 2019 did not take place. But the effort was not in vain – Lockheed Martin acquired the bankrupt company Vector, continued cooperation with USC and in January 2022 still launched Gsky-1, albeit under the name Dodona. The company positioned it as the first satellite using software-defined satellite architecture SmartSat. The system runs on the NVIDIA Jetson platform and includes the NVIDIA JetPack artificial intelligence application builder. AI systems are used to process images and digital signals.
In May 2021, San Francisco-based startup Loft Orbital was awarded a contract by the Small Business Innovation Research Program with the U.S. Space Force to develop an edge computer that can analyze data in space. Loft Orbital is offering customers the ability to place and manage payloads onboard its satellites through its Cockpit web portal, and plans to develop its own space services, which will be provided from the satellite using the SaaS model.
Florida-based OrbitsEdge proposes to use the colocation model on orbit by mounting servers in a standard 19-inch server rack with 5U of equipment space and connecting them to OrbitsEdge’s patented SatFrame Constellation space-protected bus. In essence, the standard servers will be housed in a space-based micro data center
HPE EL8000 servers will be installed aboard the first two SatFrame satellites. The company plans to start with low Earth orbit, then move to geosynchronous Earth orbit and near-lunar locations. In January 2022, OrbitsEdge signed a letter of intent with Rogue Space Systems, which will provide transportation, refueling, orbital inspection (sending a robotic spacecraft periodically) and maintenance services for the microODC for 10 years. The robots servicing the space microdata center do not need the costly environment required for human operations.
Directions for the development of space commercial data centers
- Interaction, up to the organization of separate communication channels, of cloud providers with companies launching computational workloads into space. Provision of space services from familiar to customers hyperscale clouds.
- Reducing the cost of space micro data centers, including through the use of standard commercial computing equipment.
- Reducing the cost of space services at the expense of resource sharing and software-defined architecture.
Energy must flow into any data center, and then the heat generated must be removed from the computing equipment. For data centers, it makes sense to use renewable energy sources, in particular solar panels. U.S. spacecraft have often used chemical sources, such as hydrogen-acid in the Apollo. But for data centers long in orbit, this approach is hardly feasible, especially considering that the chemical reaction goes with significant heat release, and heat removal in space is a difficult task.
Space is not suitable for traditional cooling – vacuum is an excellent heat insulator. To realize this, just think of the vacuum flasks in household thermoses. Space is empty, for classical (based on conduction and convection) cooling there are too few particles, even with low energy. And the farther away from the planet’s surface, the fewer there are. In deep space, the only way of heat transfer is radiation.
Spacecraft (SC) use special radiators that emit in the infrared range to transfer heat to the environment. They can be passive, when the radiated heat is transferred to them due to thermal conductivity of attachment units, or active, with a coolant pumped through them. The ISS uses ammonia as the external coolant and normal water as the internal coolant. The radiators (in Fig. 4 are the light sections. The color of the surface is closer to white, the more it radiates and absorbs less heat) allow the station to give off only 70 kW of heat power to the surrounding space, while being larger than the station’s living modules. And their area is comparable to the area of solar panels (dark sections).
Even for a small by modern standards data center for 200 racks the surface area of such radiators should be almost 20 times larger. It is difficult to imagine the payback of such a space monster, and the risks of collisions with other objects in orbit will increase significantly. So at least the near future belongs only to micro data centers.
For powerful spacecraft we will have to look for other ways of cooling. According to Stephan-Boltzmann law, the total emissivity of an absolutely black body is proportional to the fourth power of its temperature. Therefore, to cool the space data centers of the future, we will most likely have to develop high-hot, high-temperature radiators. At the same time, they should receive a minimum of energy from the outside. The main source of energy in orbit is the Sun, so heat-releasing radiators should be kept either parallel to the sunlight, or in the shadow.
When designing a spacecraft it is necessary to take into account the energy exchange not only of the spacecraft and space bodies, but also of all devices and equipment, as well as the orientation of satellites with respect to sources of both direct and reflected from other objects of radiation. In order to prevent one equipment from overheating the other and the third from freezing and to keep the operating temperature onboard the spacecraft, a separate service system for thermal mode support is being developed. The more computing load on board, the more energy has to be withdrawn and the more difficult it is to maintain thermal mode.
Depending on the type of orbit the temperature of the outer shell can vary from -170℃ to +120℃. Solar cells can’t always generate electricity, so enough energy storage is needed to keep the system running in the shade.
Satellites operate under harsh radiation conditions. Solar flares can disable electronics, disrupt communication channels with other satellites and with Earth. And safety should not be forgotten. A specific threat is collision with other space objects in orbit, primarily space debris. The environment of the data center should be monitored and, if necessary, the orbit should be corrected.
Technological challenges of space data centers
- Difficulty of heat removal. It can only be dissipated by radiation, which is difficult to implement at high power.
- Difficulty of calculating the orientation of the data center relative to direct and reflected radiation sources and maintaining the operating temperature onboard the spacecraft.
- Risks of collision with other space objects and space debris.
- The need to protect computing equipment from solar flares and radiation exposure.
However, the problems are solvable for small onboard data centers, as shown by an experiment with the placement of conventional commercial servers on the ISS. In February 2021, Hewlett Packard Enterprise partnered with NASA to deploy an edge data center on the International Space Station with access via the Microsoft Azure cloud. Spaceborne Computer 2 is a box of containers, each containing a HPE Proliant DL360 Gen10 server for high-performance computing and an Edgeline EL4000 GPU-equipped converged edge system. The performance of the computing complex was found to be satisfactory, but still, this edge data center operated in a relatively comfortable, human-capable environment on the ISS. On unmanned satellites, the computing workload must operate under harsher conditions
A glimpse into the future
Data center developers are looking beyond orbit and are already planning to place them on other space objects. On the moon, for starters. The American company Lonestar Data Holdings, which took part in the work on the ISS, announced in April 2022 about the creation of a series of data centers on the lunar surface. The company is contracted to fly to the first two missions to the moon and assemble its first data center there. Lonestar is developing the server, while Intuitive Machines is responsible for the Nova-C lander design and landing.
The pilot “lunar data center” is expected to be delivered to the surface of our satellite by the end of 2024 and placed near the Marius Hills in Procellarum Ocean in a robot-digged shaft. Active data exchange with Earth is not planned at the first stage – micro DPC, rather, will play the role of a backup with the most important and unchanging information of the planet. It is not clear, though, who will be able to use it, if any global cataclysm happens on Earth, when this backup is needed. In the case of successful completion of work in the future there will be a full information exchange between the Earth and the Moon.