Exploiting idle CPU cores to improve file access performance

ABSTRACT

Many-core CPUs require many parallel computation tasks to reach their full potential because CPU cores become idle if they do not have enough computation tasks. How best to utilize a number of cores in many-core CPUs should be examined. In this paper, we propose exploitation of idle cores for improving file access performance. Idle cores are used to extract file access patterns from access logs and the extracted patterns are used to improve file cache efficiency by reordering the LRU (Least Recently Used) list based on the extracted patterns. Data mining techniques are used to extract access patterns to reduce computation overhead. Our method was evaluated by simulation and also implemented on Linux kernel 2.6.26 as a prototype system. In the simulation experiment, our method improved the cache-hit ratio up to 1.09% on DBT-2 (TPC-C) trace logs. Our prototype implementation on Linux improves DBT-2 performance up to 5.24% on a real machine.

REFERENCES

• 1: Daniel Bovet , Marco Cesati, Understanding The Linux Kernel, Oreilly & Associates Inc, 2005
• 2: Gary A. S. Whittle , Jehan-François Pâris , Ahmed Amer , Darrell D. E. Long , Randal Burns, Using Multiple Predictors to Improve the Accuracy of File Access Predictions, Proceedings of the 20 th IEEE/11 th NASA Goddard Conference on Mass Storage Systems and Technologies (MSS'03), p.230, April 07-10, 2003
• 3: James Griffioen , Randy Appleton, Reducing file system latency using a predictive approach, Proceedings of the USENIX Summer 1994 Technical Conference on USENIX Summer 1994 Technical Conference, p.13-13, June 06-10, 1994, Boston, Massachusetts
• 4: Jeanna Matthews , Sanjeev Trika , Debra Hensgen , Rick Coulson , Knut Grimsrud, Intel® Turbo Memory: Nonvolatile disk caches in the storage hierarchy of mainstream computer systems, ACM Transactions on Storage (TOS), v.4 n.2, p.1-24, May 2008 [doi>10.1145/1367829.1367830]
• 5: L. Yu, G. Chen and J. Dong, "Mining Infrequently-Accessed File Correlations in Distributed File System," In Proc. of the Joint Conf. of the 9th Asia-Pacific Web Conf. and the 8th Int'l Conf. on Web-Age Information Management (APWeb/WAIM), pp.630--641, Huang Shan, China, Jun. 2007.
• 6: Steven W. Schlosser , Jiri Schindler , Stratos Papadomanolakis , Minglong Shao , Anastassia Ailamaki , Christos Faloutsos , Gregory R. Ganger, On multidimensional data and modern disks, Proceedings of the 4th conference on USENIX Conference on File and Storage Technologies, p.17-17, December 13-16, 2005, San Francisco, CA
• 7: Theodore Johnson , Dennis Shasha, 2Q: A Low Overhead High Performance Buffer Management Replacement Algorithm, Proceedings of the 20th International Conference on Very Large Data Bases, p.439-450, September 12-15, 1994
• 8: Vivekanand Vellanki , Ann L. Chervenak, A cost-benefit scheme for high performance predictive prefetching, Proceedings of the 1999 ACM/IEEE conference on Supercomputing (CDROM), p.50-es, November 14-19, 1999, Portland, Oregon, United States [doi>10.1145/331532.331582]
• 9: Xiaoning Ding , Song Jiang , Feng Chen, A buffer cache management scheme exploiting both temporal and spatial localities, ACM Transactions on Storage (TOS), v.3 n.2, p.5-es, June 2007 [doi>10.1145/1242520.1242522]
• 10: X. Yan, J. Han and R. Afshar, "CloSpan: Mining closed sequential patterns in large datasets," In Proc. of the 2003 SIAM Int'l Conf. Data Mining (SDM'03), San Francisco, US-CA, May 2003.
• 11: Youjip Won , Hyungkyu Chang , Jaemin Ryu , Yongdai Kim , Junseok Shim, Intelligent storage: Cross-layer optimization for soft real-time workload, ACM Transactions on Storage (TOS), v.2 n.3, p.255-282, August 2006 [doi>10.1145/1168910.1168912]
• 12: Zhenmin Li , Zhifeng Chen , Yuanyuan Zhou, Mining block correlations to improve storage performance, ACM Transactions on Storage (TOS), v.1 n.2, p.213-245, May 2005 [doi>10.1145/1063786.1063790]
• 13: Database Test Suite, http://osdldbt.sourceforge.net/.
• 14: Transaction Processing Performance Council, http://www.tpc.org/.

INDEX TERMS

Primary Classification:
  D. Software
  D.4 OPERATING SYSTEMS
      D.4.2 Storage Management
          Subjects: Secondary storage

Additional Classification:
  D. Software
  D.4 OPERATING SYSTEMS
      D.4.2 Storage Management
          Subjects: Storage hierarchies

General Terms:
Algorithms, Design, Experimentation, Management, Measurement, Performance

Keywords:
access pattern, access pattern mining, buffer caching, many core, replacement algorithm