High performance scientific applications in heterogeneous cloud computing environment

Ogrizović, Dario. (2016). High performance scientific applications in heterogeneous cloud computing environment. PhD Thesis. Filozofski fakultet u Zagrebu, Department of Information Science.
(Poslijediplomski doktorski studij informacijskih i komunikacijskih znanosti) [mentor Mateljan, Vladimir].

PDF (Croatian) - Registered users only Download (5MB) | Request a copy

Abstract

High performance scientific applications over the years have been executed on traditional high performance computing (HPC) systems like supercomputers and clusters. With Cloud Computing users have ability to rapidly provision and access new clusters with dynamic reallocation and efficient charge back of used computational resources. Although Cloud Computing offers great potential for scientific applications, Clouds have been designed for running business and web applications, whose resource requirements are different from high performance scientific applications which typically require low latency and high bandwidth interconnections and parallel file systems to achieve high level of performance. Specific characteristics and requirements are determined, model is made and system for execution of the high performance scientific applications in heterogeneous cloud computing is implemented. In order to determine the impact of virtualization on execution of the high performance scientific applications in cloud computing a comparison has been made between application performance and synthetic benchmarks on the physical system and the system with operating system level virtualization. Settings and the measurement results for each application and synthetic benchmark were presented.

Actions (login required)

View Item