IT4Innovations - VŠB – Technical University of Ostrava

  • Beatriz OLITE profile
    Beatriz OLITE
    23 October 2018 - updated 2 years ago
Voting is closed. 348 users have voted.
Country: 
Czech Republic

About the innovator

We are one of the teams of the “Advanced Data Analysis and Simulations Lab” at IT4Innovations national supercomputing center, which is an institute at VŠB – Technical University of Ostrava. Our research focuses on High Performance Computing (HPC) scientific pipelines and related programming models for supercomputers and future exascale systems. Such pipelines are used in various application domains such as chemoinformatics, civil engineering, traffic management. Each domain requires easy specification and execution of specialised workflow execution systems and scheduling algorithms.

VSB on Twitter

What is the innovation

HyperLoom is an open-source framework for defining and executing workflow pipelines in large-scale distributed environments. Scientific pipelines (eg. machine learning) are usually composed of multiple interconnected data processing tasks which utilize thousands of computational cores, that are available in a supercomputers, in parallel. HyperLoom users can easily define dependencies among computational tasks and create a pipeline which can then be executed on supercomputers. HyperLoom has been used in production at industrial scale.

Out of the lab. Into the market

HyperLoom has been already used on large scale machine learning pipelines for pharma industry, and even by pharma partners themselves together with scalable matrix factorization framework SMURFF developed by Imec Leuven. We also use HyperLoom in other domains represented for example by civil engineering. We provide easy access for non-HPC specialists for execution of HyperLoom pipelines on supercomputer infrastructure by specialized HPC-as-a-Service framework.

Benefits of participation in the Framework Programme

One of the primary goals of our laboratory is supporting industry, SMEs, and academia in the field of advanced data analysis and simulations. ExCAPE project helped to connect HPC specialists with domain experts (machine learning people) and with industry (AstraZeneca and Janssen), and to test drive remote execution by HPC-as-a-Service. The simple HyperLoom interface allows defining pipelines by non-HPC experts while utilising the performance of supercomputers. This brings these technologies to a broader audience and provides opportunities to increase partnerships.

This innovation was funded via H2020 project ExCAPE

Team behind the innovation

/futurium/en/file/hyperloomteamjpghyperloomteam.jpg