DART-MPI: An MPI-based Implementation of a PGAS Runtime System

TL;DR

This paper introduces DART, a PGAS runtime system built on MPI-3, enabling simplified parallel programming on large-scale clusters through one-sided communication, with performance evaluations highlighting its efficiency and limitations.

Contribution

The paper presents DART, a novel MPI-3-based PGAS runtime system that simplifies parallel application development on high-performance clusters.

Findings

DART achieves comparable performance to MPI-3 for key kernels.

Overheads are minimal for large data transfers.

Limitations identified in synchronization and small message latency.

Abstract

A Partitioned Global Address Space (PGAS) approach treats a distributed system as if the memory were shared on a global level. Given such a global view on memory, the user may program applications very much like shared memory systems. This greatly simplifies the tasks of developing parallel applications, because no explicit communication has to be specified in the program for data exchange between different computing nodes. In this paper we present DART, a runtime environment, which implements the PGAS paradigm on large-scale high-performance computing clusters. A specific feature of our implementation is the use of one-sided communication of the Message Passing Interface (MPI) version 3 (i.e. MPI-3) as the underlying communication substrate. We evaluated the performance of the implementation with several low-level kernels in order to determine overheads and limitations in comparison to the underlying MPI-3.