Keywords: grid computing, web services, grid metaschedulers, service-oriented systems, structural analysis, HPC techniques.

The use of distributed computing technologies in the field of structural analysis for large structures, because of the computational and memory requirements of the problem, provides a new variety of possibilities to be exploited. Recently, a new technology, known as Web Services, has emerged as the facto standard in Distributed Computing. Web Services are composed of XML-based protocols that offer flexible and extensible mechanisms for describing, discovering and invoking network services.

The Globus Toolkit, one of the most extended and used Grid Middleware, has adopted Web Services technology to define and establish its architecture and component interfaces in its version 4 (GT4) [1]. As a result, the so-called Grid Services have appeared as Web Services that extend their general functionality to the Grid domain, providing a standard interface to all its services, such as GRAM, MDS, etc. In this paper, a GT4-based Structural Analysis Grid Service (SAGS) for on demand 3D realistic static and dynamic linear simulation of high-rise buildings have been developed and deployed.

SAGS is a Grid Service implementation that performs a static and dynamic structural analysis by means of an HPC-based structural simulator [2]. Using Grid Services technology, and its standard XML-based protocols, the service publishes a set of methods which cover all the requirements of a reliable structural analysis. These methods are invoked via SOAP requests, enabling the usage of heterogeneous clients to interact with the service. This is an important feature, because it preserves the client preferences regarding the programming language, the execution platform, etc.

Several Grid metaschedulers such as GMarte [3], Condor-G [4] and Grid Service Broker [5] have been evaluated in order to find the most suitable one, which satisfies the requirements of high throughput, reliability and functionality.

Several fault tolerance levels have been developed to preserve the client requests as well as the service integrity. The implemented fault tolerance schema requires specific mechanisms to guarantee that any request submitted to the service will be successfully processed, considering possible failures in the network, the client, the service, the parallel application and the execution machine.

On one hand, a network or client failure during the structural analysis is handled by keeping the output of the simulation results in the service, until the client has recovered. Parallel application and execution machine failures are automatically handled by the Scheduler, re-allocating the simulation on another available resource. On the other hand, a service status persistence schema guarantees that a subsequent service start up will execute all the pending requests in case of service failure.

On the other hand, a software client, which interacts with the service via SOAP requests, has been implemented. It consists of an advanced graphical user interface (GUI) application that provides several user-friendly tools addressed to perform the pre-processing and post-processing phases. By means of Java 3D graphic library, this multiplatform application shows the structure in a highly interactive 3D scene, offering typical CAD functionalities such as rotations, translations, zooming, selections, etc., and two modes of visualisation, wired and solid. Once all the building properties have been defined by the user, the client employs the SAGS to analyse the structure, by means of its public interface. To achieve this goal, a XML file with the building features, together with different parameters related to the analysis, is sent to SAGS. Then, SAGS carries out the task simulation on a Grid infrastructure composed, in our case, of several clusters of PCs, periodically evaluating its status. Once the structural analysis has finished, the output data are retrieved by the client and deleted from the Grid Service host. Finally, the results obtained are shown, in an intuitive way, thanks to the 3D post-processing stage, offering a realistic view to evaluate them properly.

1

I. Foster, "Globus Toolkit Version 4. Software for Service-Oriented Systems", IFIP International Conference on Network and Parallel Computing, Springer-Verlag, LNCS 3779, 2-13, 2005. doi:10.1007/11577188_2

2

J.M. Alonso, V. Hernández, "Three-dimensional Structural Dynamic Analysis using Parallel Direct Time Integration Methods", The Tenth International Conference on Civil, Structural and Environmental Engineering Computing (CC 2005), paper 232. doi:10.4203/ccp.81.232

3

J.M. Alonso, V. Hernández, G. Moltó, "GMarte: Grid Middleware to Abstract Remote Task Execution", To appear in the Concurrency and Computation: Practice and Experience, 2006. doi:10.1002/cpe.1052

4

J. Frey, T. Tannenbaum, I. Foster, M. Livny, S. Tuecke. "Condor-G: A Computation Management Agent for Multi-Institutional Grids", Proceedings of the Tenth IEEE Symposium on High Performance Distributed Computing (HPDC10), 55-63, 2001. doi:10.1109/HPDC.2001.945176

5

S. Venugopal, R. Buyya, L. Winton, "A Grid Service Broker for Scheduling Distributed Data-Oriented Applications on Global Grids", Proceedings of the 2nd International Workshop on Middleware for Grid Computing, ACM Press, 2004. doi:10.1145/1028493.1028506

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £105 +P&P)