Publication: Adaptation, deployment and evaluation of a railway simulator in cloud environments
Loading...
Identifiers
Publication date
2014-06
Defense date
2014-07-14
Authors
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
Many scientific areas make extensive use of computer simulations to study realworld
processes. As they become more complex and resource-intensive, traditional
programming paradigms running on supercomputers have shown to be limited by
their hardware resources.
The Cloud and its elastic nature has been increasingly seen as a valid alternative
for simulation execution, as it aims to provide virtually infinite resources, thus
unlimited scalability. In order to bene t from this, simulators must be adapted to
this paradigm since cloud migration tends to add virtualization and communication
overhead.
This work has the main objective of migrating a power consumption railway
simulator to the Cloud, with minimal impact in the original code and preserving
performance. We propose a data-centric adaptation based in MapReduce to distribute
the simulation load across several nodes while minimising data transmission.
We deployed our solution on an Amazon EC2 virtual cluster and measured its
performance. We did the same in in our local cluster to compare the solution's performance
against the original application when the Cloud's overhead is not present.
Our tests show that the resulting application is highly scalable and shows a better
overall performance regarding the original simulator in both environments.
This document summarises the author's work during the whole adaptation development
process .
Description
Keywords
Railway simulator, Cloud computing, Simulation