High-throughput sockets over RDMA for the Intel Xeon Phi coprocessor

In this paper we describe the design, implementation and performance of Trans4SCIF, a user-level socket-like transport library for the Intel Xeon Phi coprocessor. Trans4SCIF library is primarily intended for high-throughput applications. It uses RDMA transfers over the native SCIF support, in a way...

Descripción completa

Detalles Bibliográficos
Autores principales: Santogidis, Aram, Lalis, Spyros
Lenguaje:eng
Publicado: 2017
Materias:
Computing and Computers
Acceso en línea:https://dx.doi.org/10.1007/978-3-319-67630-2_6
http://cds.cern.ch/record/2302593
_version_ 1780957289676537856
author Santogidis, Aram
Lalis, Spyros
author_facet Santogidis, Aram
Lalis, Spyros
author_sort Santogidis, Aram
collection CERN
description In this paper we describe the design, implementation and performance of Trans4SCIF, a user-level socket-like transport library for the Intel Xeon Phi coprocessor. Trans4SCIF library is primarily intended for high-throughput applications. It uses RDMA transfers over the native SCIF support, in a way that is transparent for the application, which has the illusion of using conventional stream sockets. We also discuss the integration of Trans4SCIF with the ZeroMQ messaging library, used extensively by several applications running at CERN. We show that this can lead to a substantial, up to 3x, increase of application throughput compared to the default TCP/IP transport option.
id cern-2302593
institution Organización Europea para la Investigación Nuclear
language eng
publishDate 2017
record_format invenio
spelling cern-23025932019-09-30T06:29:59Zdoi:10.1007/978-3-319-67630-2_6http://cds.cern.ch/record/2302593engSantogidis, AramLalis, SpyrosHigh-throughput sockets over RDMA for the Intel Xeon Phi coprocessorComputing and ComputersIn this paper we describe the design, implementation and performance of Trans4SCIF, a user-level socket-like transport library for the Intel Xeon Phi coprocessor. Trans4SCIF library is primarily intended for high-throughput applications. It uses RDMA transfers over the native SCIF support, in a way that is transparent for the application, which has the illusion of using conventional stream sockets. We also discuss the integration of Trans4SCIF with the ZeroMQ messaging library, used extensively by several applications running at CERN. We show that this can lead to a substantial, up to 3x, increase of application throughput compared to the default TCP/IP transport option.oai:cds.cern.ch:23025932017
spellingShingle Computing and Computers
Santogidis, Aram
Lalis, Spyros
High-throughput sockets over RDMA for the Intel Xeon Phi coprocessor
title High-throughput sockets over RDMA for the Intel Xeon Phi coprocessor
title_full High-throughput sockets over RDMA for the Intel Xeon Phi coprocessor
title_fullStr High-throughput sockets over RDMA for the Intel Xeon Phi coprocessor
title_full_unstemmed High-throughput sockets over RDMA for the Intel Xeon Phi coprocessor
title_short High-throughput sockets over RDMA for the Intel Xeon Phi coprocessor
title_sort high-throughput sockets over rdma for the intel xeon phi coprocessor
topic Computing and Computers
url https://dx.doi.org/10.1007/978-3-319-67630-2_6
http://cds.cern.ch/record/2302593
work_keys_str_mv AT santogidisaram highthroughputsocketsoverrdmafortheintelxeonphicoprocessor
AT lalisspyros highthroughputsocketsoverrdmafortheintelxeonphicoprocessor

Ejemplares similares