By Rob Mitchum // March 7, 2017
Personal computers and smartphones have democratized computing around the world, giving regular people access to technology capable of conducting billions of calculations per second. But a significant computing gap remains between consumer technology and the high-performance computers used in national laboratories and universities, and not just because of the state-of-the-art supercomputers that only these institutions can afford to run. Parallel computing — the practice of using multiple processors simultaneously to complete tasks faster and more efficiently — is now crucial for scientific computing, but the approach has still only made a light footprint on computation at home and in the office.
At the Computation Institute, CI Senior Fellow Michael Wilde leads a research staff developing Swift, a scripting language that makes parallel computing easier for scientists to deploy. But for the last two years, Wilde has also worked on introducing parallel computing to a broader audience through Parallel Works, a startup he co-founded with computational designer Matthew Shaxted. With funding from the University of Chicago Innovation Fund and space at the Polsky Exchange, Parallel Works seeks to bring cloud-based parallel features to companies in construction, finance, and energy sectors. Their work was recently featured by Chicago Inno reporter Ustav Gandhi:
Parallel Works’ platform takes supercomputing to the software-as-a-service, pay-per-use model, thus removing talent and resource maintenance costs typically associated with big machines. The software is domain agnostic–employed in applications as diverse as climate change research to neuroscience to mass media analysis–and the technology’s custom interface can be packaged, stored, and deployed as per the needs of the client or industry. Previously, with resource constraints, industry experts would have to conduct a time-consuming process of creating custom workflows that could handle large-scale parallel computing. Parallel Works hopes to make that process simpler, with a browser-based online platform and a cloud-based workflow — similar to using an app store.
Demand for these services is rising as industries start to move forward from data collection to data analysis, Wilde told the reporter. With parallel computing, companies can now begin to realize modeling and predictive analytics applications that would previously have been too slow and computationally expensive for regular use.
“Cloud computing, advances in software technology (like open source simulation tools), and new programming models now allow for advanced computational methods by companies that never had access to such techniques in the past,” said Michael, in an interview with Chicago Inno. “We are already seeing big computing as the next wave after big data.”
For more on the founding and initial funding of Parallel Works, read a CI article from 2015, or read about Swift and its applications at Argonne’s Advance Photon Source. For job postings and other information on Parallel Works, visit their website.