An easy to use distributed computing framework

Abstract

This paper introduces a client-server suite, which provides a readily configurable and simple to use network distributed computing framework. The suite will allow users to easily operate a distributed computing environment and solve real problems with a hybrid peer-to-peer (P2P) grid. This tool is most useful in a pedagogical setting such as an introductory computing course or as an adjunct to a computationally intensive course that addresses parallel computing issues.

Is teaching parallel algorithmic thinking to high school students possible?

“…As such, XMT provides students a simple-to-use alternative to MPI. As has been noted by others [3,5,8,9,10] the complexity of coding in MPI or even OpenMP can be quite overwhelming for a beginning student. With XMT the coding overhead is instead very light.…”

Cluster computing for web-scale data processing

“…We contend that traditional platforms for distributed systems programming such as MPI or PVM (as Cunha and Lourenço used) are unnecessarily complicated for a first introduction to distributed systems programming, a view corroborated by Pheatt in [10]. Pheatt specifically notes that PVM, MPI and BOINC-based clusters require high amounts of overhead.…”

“…Hadoop runs on commodity Linux clusters and presents a straightforward programming interface suitable for use by students in the time frame of an academic quarter. By contrast, the use of existing platforms such as PVM in a classroom [5] introduces a high level of overhead [10]. Furthermore, courses based on PVM or MPI then typically focus on how to implement distributed systems algorithms for correctness and reliability on top of these bases; our course instead concentrates immediately on using an existing platform to complete real-world data processing tasks.…”