ABSTRACT
Due to VFS architecture limitation, Linux NFSv4 and 4.1 client cannot join RPC requests into compounds even in cases when it is allowed by protocol specification. This leads to the high sensitivity to the network latency and loss of performance on metadata-intensive operations, especially on workloads when many small files are opened. Similar issue exists in other Unix-like kernels. We propose a modification to VFS API that resolves this issue. We have a demo implementation of modified VFS and NFS client that shows measurable improvement of latency and general throughput on synthetic metadata-intensive tests, even with standard NFS servers.
- Dmitry Irtegov, Igor Knyazev, Julia Mallaeva, Sergey Oleynikov, Michael Rootman, and Dmitry Solovyev. 2014. About one approach to building low latency network file system. In Proceedings of the 10th Central and Eastern European Software Engineering Conference in Russia (CEE-SECR '14). ACM, New York, NY, USA,, Article 2, 9 pages. Google ScholarDigital Library
- Sage A. Weil. Ceph: reliable, scalable, and high-performance distributed storage, Doctoral Dissertation, University of California at Santa Cruz Santa Cruz, CA, USA, 2007 Google ScholarDigital Library
- Torben P. Lustre File System: Demo Quick Start Guide, Lustre Group, 2009Google Scholar
- Peter Radkov, Li Yin, Pawan Goyal, Prasenjit Sarkar, Prashant Shenoy, A Performance Comparison of NFS and iSCSI for IP-Networked Storage, Proceedings of the 3rd USENIX Conference on File and Storage Technologies, March 31--31, 2004, San Francisco, CA Google ScholarDigital Library
- Chen, M., Hildebrand, D., Nelson, H., Saluja, J., Subramony, A. S. H., & Zadok, E. (2017, February). vNFS: Maximizing NFS Performance with Compounds and Vectorized I/O. In FAST (pp. 301--314). Google ScholarDigital Library
- Haynes, Tom, and David Noveck. "Network File System (NFS) version 4 Protocol." RFC 7530 (2015).Google Scholar
- Shepler, Spencer, David Noveck, and Mike Eisler. "NFS version 4 minor version 1." RFC 5661 (2010).Google Scholar
- Chen, Ming, et al. "Newer is sometimes better: An evaluation of NFSv4. 1." ACM SIGMETRICS Performance Evaluation Review. Vol. 43. No. 1. ACM, 2015. Google ScholarDigital Library
- Satran, Julian, and Kalman Meth. "Internet small computer systems interface (iSCSI)." RFC 3720 (2004). Google ScholarDigital Library
- Project OCFS2, General-purpose cluster file system. https://oss.oracle.com/projects/ocfs2/Google Scholar
- A. Tannenbaum, M. Van Steen, 2006. Distributed Systems: Principles and Paradigms (2nd Edition), Prentice-Hall, Inc. Upper Saddle River, NJ, USA. Google ScholarDigital Library
- IEEE Std 1003.1, 2016 EditionGoogle Scholar
- Gooch, Richard and Engberg, Pekka, Overview of the Linux Virtual File System, https://www.kernel.org/doc/Documentation/filesystems/vfs.txtGoogle Scholar
- PostMark Test Profile: https://openbenchmarking.org/test/pts/postmarkGoogle Scholar
Index Terms
- On one source of latency in NFSv4 client
Recommendations
About one approach to building low latency network file system
CEE-SECR '14: Proceedings of the 10th Central and Eastern European Software Engineering Conference in RussiaIn this paper we describe an approach to building low latency shared network storage based on multi-layered unification and results of testing a prototype of this type of storage.
Linux block IO: introducing multi-queue SSD access on multi-core systems
SYSTOR '13: Proceedings of the 6th International Systems and Storage ConferenceThe IO performance of storage devices has accelerated from hundreds of IOPS five years ago, to hundreds of thousands of IOPS today, and tens of millions of IOPS projected in five years. This sharp evolution is primarily due to the introduction of NAND-...
A low-latency storage stack for fast storage devices
Modern storage systems are facing an important challenge of making the best use of fast storage devices. Even though the underlying storage devices are being enhanced, the traditional storage stack falls short of utilizing the enhanced characteristics, ...
Comments