The influence of system calls and interrupts on the performance of a PC cluster using a remote DMA communication primitive

Olivier Glück 1 Jean-Luc Lamotte 2 Alain Greiner 1
1 ASIM - Architecture des Systèmes intégrés et Micro électronique
LIP6 - Laboratoire d'Informatique de Paris 6
2 ANP - Algorithmique numérique et parallélisme
LIP6 - Laboratoire d'Informatique de Paris 6
Abstract : This paper presents an efficient MPI implementation on a cluster of PCs using a remote DMA communication primitive. For experimental purposes, the MultiPC (MPC) parallel computer was used. It consists of standard PCs interconnected through a gigabit High Speed Link (HSL) network. This paper focuses on communication software layers over the HSL network. Two implementations of MPI are described. The first one uses hardware interrupts for network events signaling and system calls in the communication critical path. The second one is based on full userlevel communications. Measures show a latency of 15 µs on a Pentium II-350 with this optimized implementation. A quantitative analysis shows how system calls and interrupts impact on communication time. To tally performance in a realistic environment, experiments were run on the Gauss elimination method using a parallel implementation of a local numerical analysis computational package (CADNA).
Document type :
Conference papers
Complete list of metadatas

https://hal.archives-ouvertes.fr/hal-01545008
Contributor : Lip6 Publications <>
Submitted on : Thursday, June 22, 2017 - 1:31:41 PM
Last modification on : Thursday, March 21, 2019 - 1:00:09 PM

Identifiers

  • HAL Id : hal-01545008, version 1

Citation

Olivier Glück, Jean-Luc Lamotte, Alain Greiner. The influence of system calls and interrupts on the performance of a PC cluster using a remote DMA communication primitive. Third International Conference on Parallel and Distributed Computing Applications and Technologies (PDCAT'2002), Sep 2002, Kanazawa, Japan. pp.414-421. ⟨hal-01545008⟩

Share

Metrics

Record views

62