Benefits of Quadrics Scatter / Gather to PVFS 2

نویسندگان

  • Weikuan Yu
  • Dhabaleswar K. Panda
چکیده

Noncontiguous IO access is the main access pattern in scientific applications. We have designed an algorithm that supports zerocopy noncontiguous PVFS2 IO using a software scatter/gather mechanism over Quadrics. To investigate what impact Quadrics scatter/gather mechanism can have on PVFS2 IO accesses, in this paper, we perform an indepth evaluation of the scatter/gather mechanism. We also study how much this mechanism can improve on the performance of scientific applications. Our performance evaluation indicates that Quadrics scatter/gather is beneficial to PVFS2 IO bandwidth. The performance of an application benchmark, MPITile-IO, can be improved by 113% and 66% in terms of aggregated read and write bandwidth, respectively. Moreover, our implementation significantly outperforms an implementation of PVFS2 over InfiniBand, which does This research is supported in part by a DOE grant #DE-FC02-01ER25506 and NSF Grants #CCR0311542, #CNS-0403342 and #CNS-0509452. not support zero-copy noncontiguous IO.

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

ts of Quadrics Scatter / Gather to PVFS 2

Noncontiguous IO access is the main access pattern in scientific applications. We have designed an algorithm that supports zerocopy noncontiguous PVFS2 IO using a software scatter/gather mechanism over Quadrics. To investigate what impact Quadrics scatter/gather mechanism can have on PVFS2 IO accesses, in this paper, we perform an indepth evaluation of the scatter/gather mechanism. We also stud...

متن کامل

Supporting Efficient Noncontiguous Access in PVFS over InfiniBand

Noncontiguous I/O access is the main access pattern in many scientific applications. Noncontiguity exists both in access to files and in access to target memory regions on the client. This characteristic imposes a requirement of native noncontiguous I/O access support in cluster file systems for high performance. In this paper, we address two main issues on supporting efficient noncontiguous I/...

متن کامل

Efficient RDMA-based Multi-port Collectives on Multi-rail QsNet Clusters

Many scientific applications use MPI collective communications intensively. Therefore, efficient and scalable implementation of collective operations is critical to the performance of such applications running on clusters. Quadrics QsNet is a high-performance interconnect for clusters that implements some collectives at the Elan level. These collectives are directly used by their corresponding ...

متن کامل

Parallel and Distributed Scatter-Gather Clustering System Development Proposal

From the process of scatter-gather algorithm explained above, we can easily find the essence of the parallel version of this algorithm is the parallel clustering algorithm used in the scatter phase. Frieder, et al. implements a parallel version of the buckshot clustering algorithm [1]. Their work meets the need of the parallel scatter-gather clustering algorithm pretty well, although we can des...

متن کامل

The Impact of Hardware Gather/Scatter on Sparse Gaussian Elimination

Recent vector supercomputers provide vector memory access to "randomly" indexed vectors, whereas early vector supercomputers required contiguously or regularly indexed vectors. This additional capability, known as "hardware gather/scatter," can be used to great effect in general sparse Gaussian elimination. In this note we present some examples that show the impact of this change in hardware on...

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:

دوره   شماره 

صفحات  -

تاریخ انتشار 2005