1,720,994 research outputs found
Realizzazione e valutazione delle prestazioni di un'architettura parallela a basso costo
No full text
MGF: a Grid-enabled MPI Library
No full textComputational grids allow access to several computing resources interconnected in a distributed heterogeneous infrastructure for parallel computing. This powerful resource aggregation increases the application runtime environment complexity. A simple programming model, capable
of hiding this complexity, facilitates the use of grid technology in high-performance computing. The message passing interface can play this role and make the grid more accessible to developers with parallel programming skills.
In this paper we present MGF, a grid-enabled MPI implementation which extends the existing MPICH-G2. MGF aims are: to allow the transparent use of coupled Grid resources within the MPI library; to give programmers a detailed view of the execution system network topology;
to use the most efficient channel available for point-to-point communications and finally, to mprove collective operation efficiency by introducing a delegation mechanism
Differential Evolution as a viable tool for Satellite Image Registration
No full textThe use of Particle Swarm Optimization, a heuristic optimization technique based on the concept of swarm, is described to face the problem of classification of instances in multiclass databases. Three different fitness functions are taken into account, resulting in three versions being investigated. Their performance is contrasted on 13 typical test databases. The resulting best version is then compared against other nine classification techniques well known in literature. Results show the competitiveness of Particle Swarm Optimization. In particular, it turns out to be the best on 3 out of the 13 challenged problems
MGF: a Grid-Enabled MPI Library with a Delegation Mechanism to Improve Collective Operations
No full textA Distributed Bio-Inspired Method for Multisite Grid Mapping
Full text linkComputational grids assemble multisite and multiowner resources and represent the most promising solutions for processing distributed computationally intensive applications, each composed by a collection of communicating tasks. The execution of an application on a grid presumes three successive steps: the localization of the available resources together with their characteristics and status; the mapping which selects the resources that, during the estimated running time, better support this execution and, at last, the scheduling of the tasks. These operations are very difficult both because the availability and workload of grid resources change dynamically and because, in many cases, multisite mapping must be adopted to exploit all the possible benefits. As the mapping problem in parallel systems, already known as NP-complete, becomes even harder in distributed heterogeneous environments as in grids, evolutionary techniques can be adopted to find near-optimal solutions. In this paper an effective and efficient multisite mapping, based on a distributed Differential Evolution algorithm, is proposed. The aim is to minimize the time required to complete the execution of the application, selecting from among all the potential ones the solution which reduces the use of the grid resources. The proposed mapper is tested on different scenarios
- …
