OrbitsEdge is likely to give headaches to traditional satellite providers and offer intriguing possibilities to the growing edge computing movement. The company is offering a proprietary satellite bus designed to protect off-the-shelf rack mountable computing gear from the harsh environment of space, enabling users to tap into IT resources with low latency. It’s also likely to temporarily confuse Cubesat people with its use of “U” for rack space volume.
“What we’re looking at in the past and today, all the computers that go up on satellites are vintage tech,” said Rick Ward, Chief Technology Officer at OrbitsEdge. “There’s a tremendous amount of work on radiation hardening to make sure they work for a very long time. There’s no modern computer out in space. We’re looking to change that.”
The SatFrame 445 satellite will fly in Low Earth Orbit (LEO), providing power, thermal control/cooling, improved radiation protection and a host of communications capabilities to a standard 19 inch server rack with available space for 5U of hardware up to full-size 36 inch deep hardware. In addition, software “hardening” of devices will be necessary to compensate for radiation faults and potential damage.
Radiation is the biggest threat to computing in space as solar flares and cosmic radiation randomly zip through RAM and CPUs, with best case scenarios simple “bit flips” in memory storage or processes requiring a reboot. Physical damage to chips also occurs over time, making memory and CPUs unusable. Worse yet, devices become more vulnerable to radiation as fabrication processes get smaller. Packing more transistors onto a piece of silicon means the latest generation of chips are those most likely to be brutalized and rendered ineffective by higher radiation levels found outside of Earth’s atmosphere.
Only recently has newer off-the-shelf IT and computing hardware gone up into orbit, but experience is limited. Smaller cubesats have used cell phones due to low cost and compactness along with a low-cost/lower lifetime philosophy of 1 to 3 years in orbit, while HP Enterprise (HPE) recently launched a “supercomputer” to the International Space Station. The Spaceborne computer was built around an HPE Apollo 40-class system and used a modified Linux OS, with the computer returned to Earth after over a year of operation for teardown and fine analysis.
OrbitsEdge plans to launch a “sub 300 kilogram range” satellite as a testbed for its technologies and COTS hardware, with half-deep rack (18 inch) hardware onboard. Payloads will operate on a “day/night” cycle on the demo satellite to conserve power and manage heat, powering up when solar energy is available to run devices and shutting down when on the night side of Earth.
“Our demo mission is the smallest,” said Ward. “We’re only taking what’s essential to the mission. One of the things about high capacity computing is its very power intensive. We’re running a 1 kilowatt heater, so you have to get rid of the heat. If you want to run at night, you more than double mass take that step up,” between larger solar panels and batteries needed to provide power when the sun isn’t available.
Multiple commercial architectures will be onboard, but Ward declined to provide specifics on what gear or potential CPU types may be on board. He did, however, concur with Space IT Bridge that potential load outs could include low-end CPUs, representation for GPUs such as NVIDIA, and the latest silicon. The upside to the latest chip fabrication technologies is placing multiple cores on a single chip and the ability to monitor CPUs, shutting down one when it is damaged.
Why put computing into space in the first place? One real world application is being able to process imagery faster from other LEO satellites. Radar and visual imagery are transmitted in raw form to a ground station into the data center, consuming time and bandwidth. On-orbit processing would reduce bandwidth needs and could provide a processed image directly to an end user faster. For civilian and national defense users, faster imagery processing would be a bonanza.
Other potential applications could include any requiring extremely low latency, such as financial transactions, and any that could benefit from edge computing.
A commercial version of the OrbitsEdge satellite will have multiple radios “some talking up, some talking down, some talking sideways,” said Ward, illustrating the need to send processed data back to the ground, upward for relay through a GEO communications satellite, and to communicate with other satellites for picking up and passing along raw and processed information.
OrbitsEdge is still exploring different business models. Initial satellites may be populated with servers with users leveraging VMware to run virtual instances of the apps they need with agencies and enterprises requesting more customized hardware loads tailored to specific needs. For security and speed purposes, organizations may order (buy) dedicated satellites, but potential customers need to become comfortable with and understand the advantages to on-orbit edge computing.
CEO Barbara Stinnett says OrbitsEdge has seed funding good through 2020, staff on hand and is preparing to secure a Series A round in the first quarter of 2020, talking to a mixture of venture capitalist funds and strategic partners. There are also a series of OEM announcements in the works with more information expected to be released in the upcoming months.
“We have three markets interested, all around sustainability,” Stinnett said. Oil, gas, and water infrastructure is one sector, government the second, and life sciences/health care as the third. Being able to provide easily accessible and computing resources is of interest in multiple markets.
Stinnett would not discuss how many satellites OrbitsEdge expects to put into orbit, saying the company had looked at it and would be disclosing their plans at a future time.
Space IT Bridge finds the concept of OrbitsEdge intriguing in a couple of aspects. It brings back the age-old discussion of “Big Dumb Pipe” verses “Smart Network” started up in 1990s-2000s VON Magazine era. Big dumb pipe advocates believed if you have enough broadband and low latency, everything can be solved by hauling functions and processes back to the data center, an argument that proved significantly true with the deployment of SDN and NFV in telecom networks.
However, 5G and its introduction of edge computing has brought back discussion of a smart network. The 5G community believes edge computing is an asset in time-sensitive applications affected by latency or just waiting around for a response, but the telecom community continues to define a set of use cases where edge computing is a “win” in 5G (Due in part to the fact 5G network deployments are continuously flowing works in progress dependent on RF bandwidth dictating architecture).
LEO broadband services being deployed by OneWeb, SpaceX, Telesat, and LeoSat are the “Big Dumb Pipe” of the 2020s. Will low latency and sufficient broadband in an underserved/unserved area be good enough for many/most users and applications or will OrbitsEdge fill in the role of “Smart network” by bringing edge computing to the equation? There’s no clear answer at the moment.