Developing parallelized computing applications for cloud-based scientific computing

Abstract

In solving engineering and scientific problems using numerical methods, we often

rely on specialized numerical computing software such as Matlab. Such software

provides an ease of use that is not available in general purpose programming

languages such as Pascal or C, but comes at a high cost, low performance, and

limited flexibility. The use of the high level general purpose programming language,

Python, provides a free, easy to use, and powerful alternative to these specialized

numerical computing software.

In addition, numerical solutions to engineering and scientific problems often requires

high performance computing systems. Computing clusters of varying sizes can be

used to solve these problems, but will require the use of parallelized codes to

effectively harness the power of such computing resources. Traditional computing

clusters used within a university requires a high capital and upkeep cost, and may not

run efficiently due to the lack of sufficient workload to keep the cluster occupied.

Cloud computing services available today through the internet, provides us with near

limitless computing resources complete with storage, memory, and CPU time, on an

on-demand basis. This present us with the opportunity to dynamically assemble a

high performance computing cluster to meet any adhoc computing needs.