Microsoft’s move into space shouldn’t be surprising; it’s had an interest in providing connectivity to the whole world for a long time and has had a range of different products and services that aim to fill that need. Azure Space brings them all into one organization, supporting everything from satellite operations, ground stations and data centers, and even computer systems running on the International Space Station.
SEE: Software Installation Policy (TechRepublic)
Space is, as Douglas Adams wrote, big. That means there’s quite a big business opportunity with the current boom in space technologies thanks to low-cost launches and the increasing use of commercial off-the-shelf hardware and software in smallsats. Microsoft isn’t alone in using its public cloud as a way to deliver space projects, with Amazon also offering satellite services.
In the past Microsoft’s space efforts have been relatively small consulting exercises, working with third parties and hardware vendors, to deliver something very specific, whether a piece of custom code or modifying existing software for use in space-ready hardware. That side of the space business isn’t going away, but much of Azure Space’s focus is on delivering larger projects, either outsourcing key functionality for existing operators, or providing compute capabilities that treat satellites as part of the ever-growing edge.
With so many different services, there’s no one Azure Space customer. Some of its early partnerships were an extension of Azure’s ExpressRoute service that used satellite links for access to Azure data centers from remote locations. Using these links to geostationary satellites, Azure services could be used, albeit with some latency, from sites like fly-in, fly-out mines in Australia and Canada where fiber connectivity is unavailable.
Manage your missions
Other tools include software for planning and managing space missions. If you’re an operator planning to build a massive satellite constellation, like those used by OneWeb and SpaceX, you need software that can model the complexity of thousands of intersecting orbits, with satellite-to-ground and satellite-to-satellite communication links. Traditional satellite control systems aren’t designed to manage the scale of these new systems, so Microsoft has developed a tool, Azure Orbital Emulator, that lets operators simulate these fast-changing networks, to test satellite-based applications before they’re loaded onto the limited compute capability of small satellites.
One important use for Azure’s Orbital Emulator is the development and training of machine learning-based control systems for satellites. These can be trained in the emulator, using hyperscale cloud tools like Azure’s machine learning platform, before being exported to run on satellite-hosted inference systems. Microsoft has built a set of pre-prepared environments for the system, so that different types of satellite operator can work with sample data, including providing sample imagery so that Earth resources operators can test out on-satellite image processing applications.
It’s hard to debug code running on a satellite, with latency, limited connectivity windows, and constrained compute resources. Being able to use Azure Space for this can be a big saver, as your code can be validated and tested in advance of an expensive upload, allowing you to get it right the first time.
However, connectivity remains a key driver for the service, building on the original Azure Orbital product. Microsoft describes it as “ground station as a service,” using partnerships with SpaceX Starlink and SES to connect the Azure cloud to its containerized Mobile Data Center. With access from low earth orbit and medium earth orbit, as well as to the familiar geostationary satellites, Microsoft is aiming to ensure global coverage for satellite communications.
The Mobile Data Center is intended to bring computers to remote locations, using a satellite uplink where connectivity is poor. By running applications in the mobile data center, you can process and upload data without worrying about the latency of a satellite connection. Microsoft is working on its own Azure-hosted software-defined radio platform based on Ubuntu, which will allow you to build your own receiver software to work with satellite (and other ground-based) radio systems
Using Azure Orbital, you can bring your own ground station or take data from one of Microsoft’s, located in its own Azure data centers, connecting your existing infrastructure to satellite. This approach works well for all types of mission, not just communications. For example, if you’re working with your own smallsat Earth resources spacecraft, you can use Azure Orbital to download imagery from your satellite, before processing it in Azure. The intent is to make the satellite an extension of your Azure application, much like working with data from an Azure Stack hybrid cloud system where data is initially processed on the edge before being uploaded to Azure for further work.
Running Azure in space!
Azure Space gives Microsoft a way to deliver one-off projects for big commercial and government clients. One interesting example of this is a project for NASA, working with Hewlett Packard Enterprise to run Azure workloads on its second generation Spaceborne Computer, built around HPE’s ruggedized Edgeline systems. This uses off-the-shelf hardware with custom software to provide high-power computers to the International Space Station. The SBC-2 project is investigating whether standard computer equipment can work in space without needing significant radiation hardening, saving NASA money.
Microsoft will be delivering Azure workloads to the SBC-2 hardware, running machine learning and other pre-processing applications, before delivering data to Azure for further work. It’s not quite clear how this will work, whether Microsoft is providing a version of Azure Stack HCI or if it’s using containers to host edge workloads via tools like Azure Arc. However, it’s an interesting example of how Azure Space intends to work with its satellite customers, providing software in space and on the ground. Other Azure Space projects involve Microsoft technologies like HoloLens augmented reality and the FarmBeats precision agriculture platform.
While Azure Space may seem a little disjointed at present, with a mix of services that at first sight might not obviously connect, it’s clear that there’s a vision here. Microsoft sees space as an extension of Azure, a more distant intelligent edge, to paraphrase Satya Nadella’s catchphrase.
With Azure Space it’s providing two parts of that edge, connectivity with the ground-based cloud and a way to run Azure applications in space. At the same time, it gives Azure a way to operate where traditional connectivity isn’t available, filling a big gap in bringing compute at scale to where it’s desperately needed. What we see today is a scaffolding, with Microsoft filling out the gaps to make space a seamless part of the modern cloud.