Easy-Par: A Hybrid Environment Based on Message-Passing and Distributed Shared Memory
Mário A. R. Dantas, Alba Cristina Magalhães Alves de Melo · Kluwer Academic Publishers eBooks · 2005
In this paper we discuss concepts of an environment, which is being developed, to help programmers to implement parallel applications under the message-passing or distributed-shared memory (DSM) paradigms. The main goal of this environment, called Easy-Par, is to provide some parameters of the distributed environment to application programmers. These parameters can help programmers to decide whether an application should consider the exchange of messages between processes or should access a central pool of shared variables. An early knowledge of the number of available workstations, the homogeneity or heterogeneity of the workstations’ architcture, the load of processors, the memory utilisation, the network workload are some examples that illustrate the desirable information which can enhance the performance of parallel applications. As a result, the Easy-Par environment targets to improve the execution of an parallel application, using when appropriate the message-passing or distributed shared-memory approach, on a network of workstations.