Engineering simulations, in particular multi-physics and multi-scale whole engine simulations, such as those targeted by the ASiMoV Prosperity Partnership project, generate vast amounts of data (on the order of many TBs), which need to be read, written, passed between simulations, post-processed, stored and eventually archived - data related to aero-engine modelling is also highly sensitive must be securely stored for the lifetime of the engine model. The ASiMoV project is led by EPCC, the supercomputing centre at the University of Edinburgh.
For this PhD studentship, which is part of ASiMoV, research topics will cover the investigation of storage class memory, which acts as both memory and storage, is node-local and represents the first tier in the storage hierarchy, sitting between conventional memory and external storage: how can deep memory and storage hierarchies can be used to move extreme-scale simulation data transparently and safely through the HPC system as part of a multi-physics simulation, without impacting on the computation? Research may also investigate Hierarchical Storage Management (HSM) methods that dynamically and transparently move data through the levels of the storage hierarchy, and that allow for data to be shared between collaborators with differing data management and security requirements. Investigations may also look at whether storing data as in customised data formats and objects rather than files will improve performance, and whether POSIX compliance can be dropped for better scalability.
Informal enquiries are encouraged and should be addressed to Dr Michele Weiland ([email protected]).
Candidates must submit an application form and are expected to meet the admissions criteria of the University and of the industrial partner.