Improving MPI Support for Applications on Hierarchically Distributed Resources
نویسندگان
چکیده
Programming non-embarrassingly parallel scientific computing applications such as those involving the numerical resolution of system of PDEs using mesh based methods for grid computing environments is a complex and important issue. This work contributes to this goal by proposing some MPI extensions to let programmers deal with the hierarchical nature of the grid infrastructure thanks to a tree representation of the processes as well as the corresponding extension of collective and point-to-point operations. It leads in particular to support N ×M communications with transparent data redistribution. Key-words: MPI, Grids, tree structure, hierarchical communication, data redistribution. in ria -0 01 50 54 3, v er si on 2 31 M ay 2 00 7 Une extension MPI pour les ressources distribues hiérarchiques Résumé : La programmation des applications massivement parall éles pour le calcul scientifique sur grilles est une probl ématique complexe et importante. Un exemple de ce type d’applications est la r ésolution num érique de systèmes à équations diff érentielles partielles qui utilisent des m éthodes de maillage. Le travail pr ésent é dans ce rapport se propose d’ étendre MPI afin de permettre aux d éveloppeurs d’avoir une approche de programmation parallèle guid ée par la nature hi érarchique des infrastructures de grilles. L’approche propos ée est bas ée sur une repr ésentation des processus en arbre et une extension des communications point à point et collectives correspondantes. En particulier, le travail se base sur la mise en place de communications N ×M comprenant la redistribution transparente des donn ées. Mots-clés : MPI, Grilles, structure en arbre, communication hierarchique, redistribution de donn ées. in ria -0 01 50 54 3, v er si on 2 31 M ay 2 00 7 Improving MPI support for applications on hierarchically distributed resources 3
منابع مشابه
Building a Highly Scalable MPI Runtime Library on Grid using Hierarchical Virtual Cluster Approach
For large computational grid systems, the Message Passing Interface (MPI) is one of the most widely used programming models for distributed computing. However, current MPI implementations cannot utilize all computing nodes on a grid as a result of network configurations that effectively hide nodes behind a frontend node or gateway. This work presents an approach called Hierarchical Virtual Clus...
متن کاملA Grid based distributed simulation of Plasma Turbulence
Grid technology is widespreading, but most grid-enabled applications just exploit shared storage resources rather than computational ones, or utilize static remote allocation mechanisms of Grid platforms. In this paper the porting on a Globus equipped platform of a hierarchically distributed-shared memory parallel version of an application for particle-in-cell (PIC) simulation of plasma turbule...
متن کاملA Grid-Based Distributed SVM Data Mining Algorithm
Distribution of data and manipulation allows for solving larger problems and executing applications that are distributed in nature. In this paper we present a grid-based distributed Support Vector Machine (SVM) algorithm. The Grid is a distributed computing infrastructure that enables coordinated resource sharing within dynamic organizations consisting of individuals, in situations and resource...
متن کاملEecient S I M Ulation of Message-passing in Distributed-memory Architectures
In this thesis we propose a distributed-memory parallel-computer simulation system called PUPPET (Performance Under a Pseudo-Parallel EnvironmenT). It allows the evaluation of parallel programs run in a pseudo-parallel system, where a single processor is used to multitask the program's processes, as if they were run on the simulated system. This allows development of applications and teaching o...
متن کاملResource-Elasticity Support for Distributed Memory HPC Applications
Computer simulations are alternatives to the scientific method in domains where physical experiments are unfeasible or impossible. When the amount of memory and processing speed required is large, simulations are executed in distributed memory High Performance Computing (HPC) systems. These systems are usually shared among its users. A resource manager with a batch scheduler is used to fairly a...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2007