Cloud Crystal

Lead Participant: Zenotech Ltd

Abstract

EPIC ([https://epic.zenotech.com][0]) by Zenotech provides a simple and secure way for complex computational workflows to be allocated and run on cloud-based hardware. Unlike similar services that provide an access portal to a single supplier EPIC provides a unified access point to a growing range of on-demand resources - including Amazon AWS, Cambridge University, CFMS, Hartree, HPC Midlands, Oxford University and Oracle Bare Metal Cloud. EPIC can also access internet-connected resources including internal, research or supply-chain hardware. EPIC includes job monitoring, billing, budgeting, security, privacy, and data management functions. Each workflow task can be run on the most suitable resource, subject to availability, cost, security and data location. Arup describes EPIC as "game changing" (see [https://zenotech.com][1]). Cloud Crystal will deliver a new AI / Machine Learning based forecasting system to EPIC -- with the initial training data from the existing database -- to provide users with real-time estimates of the availability of more competitively priced options for their large computing tasks. This is a key potential enabler, as cloud hardware providers charge a premium for resource reservations and heavily discount the dynamic spot market for immediate access to unused resources. The machine learning system will be developed with the Liverpool University Institute for Risk and Uncertainty and disseminated via the Uncertainty Quantification and Management (UQ&M) SIG run by the KTN / Innovate UK. The project supports HMG strategy for digital technology and is an example of the use of AI and machine learning for engineering -- delivering against the priorities identified in the recent whitepaper "Growing the Artificial Intelligence Industry in the UK" by Professor Wendy Hall for DCMS and BEIS.

[0]: https://epic.zenotech.com
[1]: https://zenotech.com

Lead Participant

Project Cost

Grant Offer

 

Participant

Zenotech Ltd, CHEPSTOW

People

ORCID iD

Publications

10 25 50