Gravel: A Communication Library to Fast Path MPI

Authors:
Anthony Danalis;Aaron Brown;Lori Pollock;Martin Swany;John Cavazos
Affiliations:
Department of Computer and Information Sciences, University of Delaware, Newark, DE 19716;Department of Computer and Information Sciences, University of Delaware, Newark, DE 19716;Department of Computer and Information Sciences, University of Delaware, Newark, DE 19716;Department of Computer and Information Sciences, University of Delaware, Newark, DE 19716;Department of Computer and Information Sciences, University of Delaware, Newark, DE 19716
Venue:
Proceedings of the 15th European PVM/MPI Users' Group Meeting on Recent Advances in Parallel Virtual Machine and Message Passing Interface
Year:
2008

Citing 4
Cited 1

ARMCI: A Portable Remote Memory Copy Libray for Ditributed Array Libraries and Compiler Run-Time Systems

Proceedings of the 11 IPPS/SPDP'99 Workshops Held in Conjunction with the 13th International Parallel Processing Symposium and 10th Symposium on Parallel and Distributed Processing
GASNet Specification, v1.1

GASNet Specification, v1.1
RDMA read based rendezvous protocol for MPI over InfiniBand: design alternatives and benefits

Proceedings of the eleventh ACM SIGPLAN symposium on Principles and practice of parallel programming
High performance RDMA protocols in HPC

EuroPVM/MPI'06 Proceedings of the 13th European PVM/MPI User's Group conference on Recent advances in parallel virtual machine and message passing interface

Protocols for wide-area data-intensive applications: design and performance issues

SC '12 Proceedings of the International Conference on High Performance Computing, Networking, Storage and Analysis

Quantified Score

Hi-index	0.00

Visualization

Abstract

Remote Direct Memory Access (RDMA) technology allows data to move from the memory of one system into another system's memory without involving either one's CPU. This capability enables communication-computation overlapping, which is highly desirable for addressing the costly communication overhead in cluster computing. This paper describes the consumer-initiated and producer-initiated protocols of a companion library for MPI called Gravel. Gravel works in concert with MPI to achieve increased communication-computation overlap by separating the meta-data exchange from the application data exchange, thus allowing different communication protocols to be implemented at the application layer. We demonstrate performance improvements using Gravel for a set of communication patterns commonly found in MPI scientific applications.