Head over to our on-demand library to view sessions from VB Transform 2023. Register Here
Microsoft’s quantum team says it’s developed an optimizing solver to address pressing scheduling problems of the deep space kind. The solver, which runs on the Azure Quantum cloud service, shows how programmers can adapt techniques from quantum computing research to run today on classical computer systems, its creators said. Microsoft marks it as part of its “Quantum Inspired Optimization Solution.”
“Quantum hardware isn’t yet available at the scale needed to actually solve real-world problems. But classical computers are,” said Anita Ramanan, senior software engineer of Azure Quantum at Microsoft. “We can develop algorithms that we call ‘quantum-inspired’ that emulate physical processes like quantum tunneling, for example, to tackle certain optimizations scenarios.”
Solvers like these can run on the Azure Quantum cloud on a variety of classical hardware, such as CPUs, GPUs, and FPGAs. Scenarios they seek to address include multivariate optimization problems where the problem space is, well, deep.
It’s not a simple problem that the deep space network job scheduler seeks to solve. Communications scheduling with missions like the Mars 2020 Perseverance Rover, the James Webb Space Telescope, and others, strains the Deep Space Network (DSN) used for data communications.
VB Transform 2023 On-Demand
Did you miss a session from VB Transform 2023? Register to access the on-demand library for all of our featured sessions.
“There are actually more requests to communicate in a given week than there are slots to fulfill those requests. It’s a very, very oversubscribed service,” Ramanan said in an interview with VentureBeat.
The variables are complex. Communications are handled by antenna stations – located in Spain, California, and Australia — that triangulate the Earth. These links can be blocked by the curvature of our planet as it rotates. So, schedules must carefully map with the movement of our planet. Also variable: the travel times of signals from the radio antennas to deep space probes.
The scheduling task for Jet Propulsion Lab workers that manage the DSN has been challenging since its first inception in the early 1960s. Today, the data loads are only growing, so scheduling solvers become ever more critical.
Azure Quantum and job scheduling
This class of solvers has potential use in many settings that need advanced job-shop scheduling, Ramanan noted. Healthcare staffing is one, supply chain and logistics scheduling are others. As COVID-19 pandemic news headlines remind us daily, scheduling and optimization problems in these areas have become acute on a local and global scale.
“Broadly, quantum-inspired optimization techniques can be used for a wide range of problems — everything from financial portfolio management and basic management to logistics, fleet management, and circuit fault diagnosis,” she said.
Microsoft work on optimization for space mission communications was done on a limited feature set, as described on its Microsoft Azure Quantum blog. At the outset, the Azure Quantum team measured runtimes of two hours or more to produce a schedule. By applying quantum-inspired optimization algorithms, the time was reduced to 16 minutes.
Some customizations — for example, code tailored to the underlying CPUs — reduced time to produce a schedule to about two minutes. A future goal is to incorporate a broader set of requirements.
Space is the place for optimization
The Deep Space Network project proved a good testing ground for Azure Quantum, according to Ramanan. It’s a good example of how the cloud service can solve large-scale problems, she said.
Yet some of this was enabled by common everyday tooling – think Python running on the cloud. Nevertheless, major effort was involved in addressing what Ramanan called “one of the largest problems we’ve seen.”
“It was an interesting challenge to get that up and running efficiently on the Quantum Inspired Optimization service,” she said. “It enabled us to improve the service for future customers.”
There were other plusses, of course – the coolness factor is high.
“It was really, really cool working with JPL as well,” added Ramanan, who – like more than just a few of us – admits to having wanted as a kid to be an astronaut.
VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.