NEW MADELEINE: a Fast Communication Scheduling Engine for High Performance Networks

Aumage, Olivier; Brunet, Élisabeth; Furmento, Nathalie; Namyst, Raymond

doi:10.1109/ipdps.2007.370476

Cited by 28 publications

(29 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It has been implemented in the PI-OMAN I/O Manager [13]. The mechanism of task offloading is exploited by the NEWMADELEINE communication library [6].…”

Section: Methodsmentioning

confidence: 99%

“…Having a global view of communication flows also permits to arbitrate access to the NICs. We showed that, instead of passing application messages to the NICs directly, buffering packets and applying optimizations improve throughput and avoid NICs saturation [6].…”

Section: A Concurrent Communicationmentioning

confidence: 99%

“…Our communication library for high performance networks is called NEWMADELEINE [6]. It is available over MX/Myrinet, Verbs/InfiniBand, Elan/QsNet and TCP/Ethernet.…”

Section: B Newmadeleinementioning

confidence: 99%

See 2 more Smart Citations

A scalable and generic task scheduling system for communication libraries

Denis

2009

2009 IEEE International Conference on Cluster Computing and Workshops

View full text Add to dashboard Cite

“…It has been implemented in the PI-OMAN I/O Manager [13]. The mechanism of task offloading is exploited by the NEWMADELEINE communication library [6].…”

Section: Methodsmentioning

confidence: 99%

Section: A Concurrent Communicationmentioning

confidence: 99%

See 1 more Smart Citation

A scalable and generic task scheduling system for communication libraries

Denis

2009

2009 IEEE International Conference on Cluster Computing and Workshops

View full text Add to dashboard Cite

“…We have implemented our proposed checksum mechanisms in NEW-MADELEINE [1] since it was more convenient for us to work in our own communication library. However, these mechanisms are intended to be generic and not specific to NEWMADELEINE, thus they could probably be implemented in any other MPI implementation.…”

Section: Related Workmentioning

confidence: 99%

“…It applies optimization strategies inbetween in order to use more efficiently the network [1]. For instance, multiple small messages from the application may be aggregated and sent as a single packet on the network.…”

Section: Amortizing the Cost Of Checksum Computationmentioning

confidence: 99%

High Performance Checksum Computation for Fault-Tolerant MPI over Infiniband

Denis

Ishikawa

2012

Recent Advances in the Message Passing Interface

View full text Add to dashboard Cite

To cite this version:Alexandre Abstract. With the increase of the number of nodes in clusters, the probability of failures and unusual events increases. In this paper, we present checksum mechanisms to detect data corruption. We study the impact of checksums on network communication performance and we propose a mechanism to amortize their cost on InfiniBand. We have implemented our mechanisms in the NEWMADELEINE communication library. Our evaluation shows that our mechanisms to ensure message integrity do not impact noticeably the application performance, which is an improvement over the state of the art MPI implementations.

show abstract

DiPOSH: A portable OpenSHMEM implementation for short API‐to‐network path

Coti

Malony

2021

Concurrency and Computation

View full text Add to dashboard Cite

Summary In this article, we introduce DiPOSH, a multi‐network, distributed implementation of the OpenSHMEM standard. The core idea behind DiPOSH is to have an API‐to‐network software stack as slim as possible, in order to minimize the software overhead. Following the heritage of its non‐distributed parent POSH, DiPOSH's communication engine is organized around the processes' shared heaps, and remote communications are moving data from and to these shared heaps directly. This article presents its architecture and several communication drivers, including one that takes advantage of a helper process, called the Hub, for inter‐process communications. This architecture allows use to explore different options for implementing the communication drivers, from using high‐level, portable, optimized libraries to low‐level, close to the hardware communication routines. We present the perspectives opened by this additional component in terms of communication scheduling between and on the nodes. DiPOSH is available at https://github.com/coti/DiPOSH.

show abstract

NEW MADELEINE: a Fast Communication Scheduling Engine for High Performance Networks

Cited by 28 publications

References 6 publications

A scalable and generic task scheduling system for communication libraries

A scalable and generic task scheduling system for communication libraries

High Performance Checksum Computation for Fault-Tolerant MPI over Infiniband

DiPOSH: A portable OpenSHMEM implementation for short API‐to‐network path

Contact Info

Product

Resources

About