Quantum computing start-up Agnostiq has produced a version of its quantum and high-performance computing middleware platform Covalent for AWS. It is designed to make it easier to move and spread processing loads between different machines and platforms.
The open-source workflow orchestration tool allows engineers, students, data scientists and developers to communicate and spread workflow as required between different processors. For example, if one quantum computer isn’t performing as required they can switch to another that is available via the AWS cloud.
Future versions of Covalent will be cloud-neutral and multi-cloud, allowing users to run operations across machines connected to different clouds or use Google batches.
Quantum computing has the ability to speed up computations when it reaches maturity over the next decade, accelerating innovations across a range of industries from finance to manufacturing but there are no consistent approaches, and relatively few machines to work on, making development difficult.
Opening quantum to enterprise on AWS
While some enterprise companies have started dabbling with quantum, including through offerings on cloud platforms such as Azure and AWS, it remains largely inaccessible to the enterprise. This is despite experts warning executives to take the impact seriously. A Harvard Business Review article from this year claims that “Quantum computing will enable businesses to better optimise investment strategies, improve encryption, discover products, and much more.”
Dr. Will Cunningham, Head of Quantum Software at Agnostiq, says enterprise has struggled to engage with quantum due to the novelty of the technology and the high level of expertise required to build applications. One of the big problems is finding people able to understand both cloud and quantum computing.
“How do you get different teams to communicate clearly with each other? This is a deep-tech field with people and academics that know nothing about the cloud, but all quantum computers today are accessed via the cloud and the cloud engineers know nothing about quantum,” he told Tech Monitor.
“We are looking at end of the decade for wider spread adoption for a brand-new technology opening a new paradigm of computing. As soon as there is a quantum computer that can do something useful every major bank, every major pharmaceutical company, everyone is going to want it.”
Tools like Covalent break workflows down into modular Python components, allowing users to easily reproduce repetitive code and avoid costly re-runs, as running experiments on quantum and supercomputers is “extremely expensive”.
“There is a central server that manages all the users’ workflows,” said Cunningham. “Within the workflow, you have different sets of tasks, some of them dependent on each other. Each of those tasks can be deployed to hardware wherever it is available, so the end-user is only thinking about how it will run, keeping costs in check and complete in a reasonable amount of time.”
Distributed computing
The Toronto-based company says its tool can act as a single-entry point for users wanting to access quantum processors, CPUs, GPUs and even quantum-inspired hardware that isn’t a true quantum computer without additional set-up or research. It comes equipped with a task scheduler that can automatically select the best hardware resource for the task and does so with a mixture of predefined and user-defined constraints.
Being able to switch load and use compute power independent of location is vital as supercomputers and quantum computers switch to more federated models, it also makes it easier for enterprise customers to manage expense when using expensive hardware like quantum computers.
“One of the big trends we see in HPC now is supercomputers are hitting a limit of how big they can scale and speaking to people at US National Labs, we hear that the vision is to give federated access to these devices. We are now at a point where geographically distributed computing will become the norm but there are challenges to operate in that kind of environment,” said Cunningham.
“You might have access to three clusters but queue time spikes on one and you need to redeploy work to another one but there are challenges with data fragmentation and code versions which take up time for doing novel research,” he added.
The tool previously worked for users on their own on-premises high-performance clusters but with a few extra lines of code it can now be set up as a cloud-based HPC environment on AWS and open up quantum computing hardware – which is in more demand than supercomputers.
“I think we will continue to see a very distributed model of computation. Quantum computers are so scarce right now that anybody could bring online a 5 qubit quantum computer and if you can connect it to the cloud and charge money people will use it,” said Cunningham, adding that queue times are half an hour or more, compared to minutes for classical resources.
Covalent supports AWS Lambda, Amazon ECS, Amazon EC2, AWS Batch, and others, with plans to roll out to Google Cloud in the future. Those who prefer to keep their workloads on-premises can now use it to “burst” into the cloud when extra computing resources are required.
“AWS does offer a huge variety of compute resources on the classical and quantum side but we don’t see Covalent as limited to AWS, we really see it as a tool to deliver compute wherever it is available. In future a batch computing user will be able to say I want to do batch computing in North America, don’t care which cloud provider or data centre just send it to whoever is available.”
Read more: Could quantum computing make our energy grid sustainable?
