Using a Virtual Machine Environment to Support Simulation on the Grid

Edoardo Pignotti (1), Gary Polhill (2), and Peter Edwards (1)

(1) School of Natural & Computing Sciences, University of Aberdeen. AB24 5UE, UK. {e.pignotti, p.edwards}@abdn.ac.uk

(2) Macaulay Land Use Research Institute, Craigiebuckler, Aberdeen. AB15 8QH, UK. g.polhill@macaulay.ac.uk

TITLE: Using a Virtual Machine Environment to Support Simulations on the Grid

ABSTRACT: Agent-based social simulation has been around for a number of years now, and whilst some people still prefer to build their own models from scratch, there are commonly used simulation modelling libraries and tools that are designed to assist social scientists in constructing such models. The most popular of these are probably RePast [6], Swarm [4], MASON [3], and for relatively simple models, NetLogo [7], but numerous others are also used [5].

While agent-based modellers differ in their approach to using simulation models, it is reasonably common practice to conduct experiments with them by varying parameters to explore parameter sets, and/or examine stochastic variation in outcome. Such a computational task is ‘embarrassingly parallel’ in that there is no need for each run of the model to communicate with any of the other runs. It would therefore seem appropriate to explore how such models could be used in the context of a distributed ‘eScience’ infrastructure.

The vision of eScience is to facilitate large scale science using Grid technologies [1]. The National Grid Service1 (NGS) is the UK’s first production level Grid for eScience and provides computing and data resources for UK researchers. Running simulation environments using eScience infrastructure is, however, a non-trivial task. The software stack necessary to run most simulation environments can be complex to install and configure Moreover, NGS resources are provisioned with a very specific OS configuration and the predefined software portfolio installed on NGS resources is narrow and general purpose. On-boarding a custom software system (e.g. Swarm) onto a set of heterogeneous NGS resources is complex and beyond the capabilities of most ‘everyday’ researchers.

We have investigated the use of virtualization technologies to enable a customized virtual machine image to run large-scale experiments on many different resources at local institutions, on the NGS or other resource/cloud providers such as EC22 or flexiscale3. This has resulted in the creation of a custom virtual machine image (simulationBox) that can be used in combination with existing eScience infrastructure (EUCALYPTUS4) to facilitate such deployment. The simulationBox image is based on a lightweight version of the Centos5.25 operating system and pre-configured with popular simulation environments (Swarm, Mason and RePast). We have also implemented mechanisms to facilitate the use of simulationBox via the Kepler [2] workflow tool. Figure 1 shows a workflow representing a simulation experiment with the FEARLUS land-use model (itself developed using Swarm). We have created a website (www.simulationbox.net) in order to make simulationBox available to the user community including documentation and tutorials describing how to use it with the UK NGS.

Acknowledgment. Funded under the JISC ENGAGE (Engaging research with e-infrastructure) programme.

References

1. Foster I, Kesselman C, Tuecke S (2001) The anatomy of the grid: Enabling scalable virtual organizations. International J. Supercomputer Applications 15(3) 2001

2. Ludascher B, Altintas I, Berkley C, Higgins D, Jeager E, Jones M, Lee E, and Tao J (2006) Scientific workflow management and the kepler system. Concurrency and Computation: Practice and Experience, pages 1039–1065, 2006.

3. Luke S, Cioffi-Revilla C, Panait L, and Sullivan K(2004) MASON: A New Multi-Agent Simulation Toolkit. Proceedings of the 2004 SwarmFest Workshop.

4. Minar, N., R. Burkhart, C. Langton, and M. Askenazi. 1996. The Swarm simulation system: A toolkit for building multi-agent simulations. Working Paper 96-06-042, Santa Fe Institute, Santa Fe.

5. Nikolai C and Madey G (2009) Tools of the trade: A survey of various agent based modeling platforms. Journal of Artificial Societies and Social Simulation 12(2): 2. http://jasss.soc.surrey.ac.uk/12/2/2.html.

6. North MJ, Collier NT and Vos JR (2006) Experiences creating three implementations of the Repast agent modeling toolkit. ACM Transactions on Modeling and Computer Simulation 16(1): 1-25.

7. Wilenski U (1999) NetLogo. htto://ccl.northwestern.edu/netlogo. Center for Connected Learning and Computer-Based Modeling. Northwestern University, Evanston, IL.

(1) www.grid-support.ac.uk; (2) aws.amazon.com/ec2; (3) www.flexiscale.com; (4) www.eucalyptus.com; (5) http://www.centos.org/)

Back to Swarmfest2010_program