Analyzing MPI performance over 10-Gigabit ethernet

نویسندگان

  • Justin Gus Hurwitz
  • Wu-chun Feng
چکیده

Recent work with 10-Gigabit (10GbE) network adapters has demonstrated good performance in TCP/IP-based localand wide-area networks (LANs and WANs). In the present work we present an evaluation of host-based 10GbE adapters in a system-area network (SAN) in support of a cluster. This evaluation focuses on the performance of the message-passing interface (MPI) when running over a 10GbE interconnect. We find that MPI over 10GbE provides communications performance comparable to that of TCP alone and fairly competitive with more exotic technologies such as MPI over Quadrics. The optimization of MPI and MPI-based applications to make use of this performance, however, is a non-trivial task. Consequently, it is difficult for MPI-based applications to realize this performance when running current-generation 10GbE hardware. Published by Elsevier Inc.

برای دانلود رایگان متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Performance Comparison of LAM/MPI, MPICH, and MVICH on a Linux Cluster Connected by a Gigabit Ethernet Network

We evaluate and compare the performance of LAM, MPICH, and MVICH on a Linux cluster connected by a Gigabit Ethernet network. Performance statistics are collected using NetPIPE which show the behavior of LAM/MPI and MPICH over a gigabit network. Since LAM and MPICH use the TCP/IP socket interface for communicating messages, it is critical to have high TCP/IP performance. Despite many efforts to ...

متن کامل

USENIX Association Proceedings of the 4 th Annual Linux Showcase

We evaluate and compare the performance of LAM, MPICH, and MVICH on a Linux cluster connected by a Gigabit Ethernet network. Performance statistics are collected using NetPIPE which show the behavior of LAM/MPI and MPICH over a gigabit network. Since LAM and MPICH use the TCP/IP socket interface for communicating messages, it is critical to have high TCP/IP performance. Despite many efforts to ...

متن کامل

USENIX Association

We evaluate and compare the performance of LAM, MPICH, and MVICH on a Linux cluster connected by a Gigabit Ethernet network. Performance statistics are collected using NetPIPE which show the behavior of LAM/MPI and MPICH over a gigabit network. Since LAM and MPICH use the TCP/IP socket interface for communicating messages, it is critical to have high TCP/IP performance. Despite many e↵orts to i...

متن کامل

Evaluation of RDMA Over Ethernet Technology for Building Cost Effective Linux Clusters

Remote Direct Memory Access (RDMA) is an effective technology for reducing system load and improving performance. Recently, Ethernet offerings that exploit RDMA technology have become available that can potentially provide a high-performance fabric for MPI communications at lower cost than other competing technologies. The goal of this paper is to evaluate RDMA over gigabit Ethernet (ROE) as a ...

متن کامل

Network Performance in High Performance Linux Clusters

Linux-based clusters have become more prevalent as a foundation for High Performance Computing (HPC) systems. With a better understanding of network performance in these environments, we can optimize configurations and develop better management and administration policies to improve operations. To assist in this process, we developed a network measurement tool to measure UDP, TCP and MPI commun...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

عنوان ژورنال:
  • J. Parallel Distrib. Comput.

دوره 65  شماره 

صفحات  -

تاریخ انتشار 2005