Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.800201
Title: Harnessing emerging supercomputers for remote and interactive visual discovery in astronomy
Author: Dykes, Timothy
ISNI:       0000 0004 8507 9781
Awarding Body: University of Portsmouth
Current Institution: University of Portsmouth
Date of Award: 2018
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
Scientific visualisation can be an instrumental data analysis tool to effectively extract meaningful information from big data generated via experimentation and observation. Visual analysis can provide a fast and intuitive aid to data comprehension that compliments traditional statistical methods. However, the rate of scientific data production is currently growing well beyond the capabilities of today’s processing pipelines, and this situation is expected to be amplified in the coming decades due to the next generation of experimentation and observation facilities and high performance computing systems. The increased size, variety, and complexity of such data presents significant challenges for scientific visualisation, and motivates creation of new and innovative tools to cope with the demands of next-generation research environments in scientific fields such as Astronomy, which is at the forefront of dataintensive disciplines with overwhelming amounts of scientific data both captured by instrumentation and generated by large-scale numerical simulations. The work presented in this thesis showcases an evolution of an existing tool called Splotch for high performance scientific visualisation, inspired by four core research topics. (1) Physical realism for particle-based volume rendering, in particular considering visual representations for multi-frequency astronomical data and synergies between astronomy and computer graphics algorithms for radiative transport. (2) Optimised utilisation of emerging architectures in modern supercomputing, specifically addressing the growing popularity of accelerator and many-core hardware platforms. (3) Interoperability with modern web-based infrastructures, focusing on a library developed to support integration of interactive high performance computing applications with common web environments for remote and interactive functionalities. (4) Integration within common scientific work-flows demonstrating the role of 3d visualisation in providing appropriate supporting mechanisms for data access and analysis in web portals. The thesis demonstrates improved quality for astronomical particle rendering, showcasing the application of an enhanced optical model within a novel pipeline for galaxy modelling and visualisation, further laying the groundwork for extended physical realism through coupling with astronomical simulation techniques. Heterogeneous high performance hardware trends are identified, and appropriate optimisation techniques are presented for many-core and graphics processing units, widening the scope of portability to high performance architectures through kernelspecific and overall efficiency improvements, and outlining methodologies applicable to next-generation architectures. A general approach for integrating high performance computing applications and web-environments is introduced as the foundation for web-based, remote, and interactive functionality in astronomical particle visualisation, and illustrated within the Theoretical Astrophysical Observatory. Future research directions are outlined regarding illumination models, galaxy modelling, performance portability, web visualisation and knowledge transfer beyond astronomy.
Supervisor: Krokos, Mel Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.800201  DOI: Not available
Share: