PDF, 2.4 MB
Zipped PostScript, 5.6 MB
PDF, 2.4 MB
Zipped PostScript, 5.6 MB
HTML
PDF, 268 KB
Zipped PostScript, 360 KB
Even though Object Orientation has been proven to be an effective programming paradigm for software development, it has not been shown to be an ideal solution for the development of large scale parallel and distributed systems. There are a number of reasons for this: the parallelism and synchronisation in these systems has to be explicitly managed by the programmer; few Object Oriented languages have implicit support for Garbage Collection in parallel applications; and the state of a systems of concurrent objects is difficult to determine. In contrast, the Condensed Graph model provides a way of explicitly expressing parallelism but with implicit synchronisation; its implementation in the WebCom system provides for automatic garbage collection and the dynamic state of the application is embodied in the topology of the Condensed Graph. These characteristics free programmers from the difficult and error prone process of explicitly managing parallelism and thus allows them to concentrate on expressing a solution to the problem rather than on its low level implementation. Object Oriented Condensed Graphs is a computational paradigm which combines Condensed Graphs with object orientation and this unified model leverages the advantages of both paradigms. This paper illustrates the basic features of Object Oriented Condensed Graphs as well as its support for large scale software development.
PDF, 471 KB
Zipped PostScript, 3.0 MB
Computational models of natural disasters are invaluable means of disaster reconstruction, crisis management, and disaster prevention. The Medigrid project addressed these issues by developing a framework of simulation models developed in previous projects. The paper describes the models incorporated into the project, together with the visualization service developed for presentation of models' outputs.
PDF, 630 KB
Zipped PostScript, 560 KB
The Grid community has made an important effort in developing middleware to provide different functionalities, such as resource discovery, resource management, job submission or execution monitoring. As part of this effort this paper addresses the design and implementation of an architecture (CPPC-G) based on services to manage the execution of fault tolerant applications on Grids. The CPPC (Controller/Precompiler for Portable Checkpointing) framework is used to insert checkpoint instrumentation into the code of sequential and MPI applications. Designed services will be in charge of submission and monitoring of the execution of CPPC-instrumented applications, management of checkpoint files generated by the fault-tolerant applications, and detection and automatic restart of failed executions.
PDF, 318 KB
Zipped PostScript, 458 KB
Recent software engineering concepts, like software as a service, allow the extension of the legacy code lifetime and the reduction of software maintenance costs. In particular, exposing computer algebra systems as services allows not only the integration in complex service-oriented architectures but also their further development. While existing standards may be used for service deployment, discovery and interaction, the particularities of services to be built require specialized solutions. A recent technical approach aimed at integrating legacy computer algebra systems into modern service-oriented architectures is presented and discussed in detail in this paper. A special emphasis is put on the ability to compose symbolic services in complex computing scenarios. A short description of how such systems were extended to allow the access of external services is provided as well. The proposed approach was implemented into a specific framework, namely, SymGrid-Services. Simple examples are provided to demonstrate usefulness of the framework.
PDF, 489 KB
Zipped PostScript, 1.1 MB
In this work MASIPE, a tool for monitoring parallel applications, is presented. MASIPE is a distributed tool that gives support to user-defined mobile agents, including functionalities for creating and transferring these agents through different compute nodes. In each node, the mobile agent can access the node information as well as the memory space of the parallel program that is being monitored. In addition, MASIPE includes functionalities for managing and graphically displaying the agent data. In this work, its internal structure is detailed and an example of a monitored scientific application is shown. We also perform a study of the MASIPE requirements (in terms of CPU and memory) and we evaluate its overhead during the program execution. Experimental results show that MASIPE can be efficiently used with minimum impact on the program performance.
PDF, 250 KB
Zipped PostScript, 463 KB
The real and specific problem that underlies the Grid concept is coordinated resource sharing and problem solving in dynamic, multi-institutional virtual organizations. This concept is realized in the most popular problem solving environments (PSEs) such as NetSolve and WebPDELab. These systems use some approaches to computational Grids and Web browser interfaces to back-end computing resources. The aim of our work is to build PSE implemented as a Web portal that allows clients to choose the most appropriate services to solve some problems using of matrix-algebra and numerical methods based on MPI techniques. In addition, it is available to extend the library of the Web-portal by loading computational algorithms. The proposed system allows to users a rapid prototyping of ideas, detailed analysis, and higher productivity.
PDF, 335 KB
Zipped PostScript, 1.0 MB
This paper presents measured communication and computational performance results on an ad-hoc computing network composed of several geographically distributed nodes. Communication speed was measured for three basic communication patterns commonly used in parallel and distributed applications: point-to-point, bidirectional ring, and all-to-all communication. For that purpose simple test programs were made, based on the MPI library. The parallel execution time of two real-life computationally intensive parallel applications is also presented. Results show that only the applications with low communication requirements could be beneficially executed on computing networks.
HTML
