Elizabeth Shriver
John Wilkes
ABSTRACT
Modern disk drives read-ahead data and reorder incoming requests in a workload-dependent fashion. This improves their performance, but makes simple analytical models of them inadequate for performance prediction, capacity planning, workload balancing, and so on. To address this problem we have developed a new analytic model for disk drives that do readahead and request reordering. We did so by developing performance models of the disk drive components (queues, caches, and the disk mechanism) and a workload transformation technique for composing them. Our model includes the effects of workload-specific parameters such as request size and spatial locality. The result is capable of predicting the behavior of a variety of real-world devices to within 17% across a variety of workloads and disk drives.
- 1.C. Adams, E. Gelenbe, and J. Vicard. An experimentally validated model of the paging drum. Acta Inforrnatica, 11:103 17, 1979.Google Scholar
- 2.R. Barve, E. Shriver, P. B. Gibbons, B. K. Hillyer, Y. Matias, and J. S. Vitter. Modeling and optimizing I/O throughput of multiple disks on a bus. A CM Conf. on Meas. and Modeling of Comp. Sys. (SIGMETRICS)/Perf. '98 (Madison, WI), June 1998. Extended abstract. Google ScholarDigital Library
- 3.A. L. Bastian. Cached DASD performance prediction and validation. Proc. 13th Intl Conf. on Mgmt. and Perf. Eval. of Comp. Sys. (CMG) (San Diego, CA), pages 174-7, M. Boksenbaum, G. W. Dodson, T. Moran, C. Smith, and H. P. Artis, editors, 14-17 Dec. 1982.Google Scholar
- 4.A. K. Bhide, A. Dan, and D. M. Dias. A simple analysis of the LRU buffer policy and its relationship to buffer warm-up transient. Proc. 9th lntl Conf. on Data Eng., pages 125-33, 23 Apr. 1993. Google ScholarDigital Library
- 5.G. Biagini. Evaluating I/O subsystem performance. Proc. Intl Conf. for the Meas. and Perf. Eval. of Comp. Sys. (CMG'86) (Las Vegas, NV), pages 299-306. Comp. Meas. Group, 1989.Google Scholar
- 6.D. Buck and M. Singhal. An analytic study of caching in computer-systems. J. of Par. and Distrib. Cornpt., 32(2):205-14, Feb. 1996. Erratum published in volume 34(2):233, May 1996. Google ScholarDigital Library
- 7.S. C. Carson and S. Setia. Analysis of the periodic update write policy for disk cache. IEEE 7Yans. on Softw. Eng., 18(1):44-54, Jan. 1992. Google ScholarDigital Library
- 8.S. Chen and D. Towsley. The design arid evaluation of RAID 5 and parity striping disk array architectures. J. of Par. and Distrib. Compt., 17(1-2):58-74, Jan.-Feb. 1993. Google ScholarDigital Library
- 9.E. G. Coffman, Jr and M. Hofri. On tile expected perfbrmance of scanning disks. SIAM J. on Computing, 11(1):60- 70, Feb. 1982.Google ScholarCross Ref
- 10.G. R. Ganger. System-oriented evaluation of I//0 subsystern performance. PhD thesis, published as Technical report CSE-TR-243-95. Dept of Comp. Science and Eng., Univ. of Michigan, June 1995. Google ScholarDigital Library
- 11.R. Golding, E. Shriver, 2'. Sullivan, and J. Wilkes. Attributemanaged storage. Workshop on Modeling and Specification of I/O (San Antonio, TX), 26 Oct. 1995.Google Scholar
- 12.C. C. Gotlieb and G. It. MacEwen. Performance of movablehead disk storage systems. J. of the A CM, 20(4):604-2:1, Oct. 1973. Google ScholarDigital Library
- 13.P. G. Hoel, S. C. Port, and C. J. Stone. Introduction to p#vbability theory. Houghton Mifflin Company, 1971.Google Scholar
- 14.M. ttofi'i. Disk scheduling: FCFS vs. SSTF revisited. Communications of the ACM, 23(11):645-53, Nov. 1980. Google ScholarDigital Library
- 15.A. Hospodor. Mechanical access time calculation. Advances in Information Storage Systems, 6:313-36, 1995.Google ScholarCross Ref
- 16.D. M. Jacobson and J. Wilkes. Disk scheduling algorithms based on rotational position. Technical report HPL-CSP- 91--7. Itewlett-Packard Labs., Palo Alto, CA, 24 Feb. 1991, revised 1 Mar. 1991.Google Scholar
- 17.L. Kleinrock. Queueing systems volume I: theory. John Wiley and Sons, 1975. Google ScholarDigital Library
- 18.D. Kotz, S. B. Toh, and S. Radhakrishnan. A detailed simulation model of the lip 97560 disk drive. #I#chnical report PCS-TR94-220. Dept of Comp. Science, Dartmouth College, Nit, 18th July 1994. Google ScholarDigital Library
- 19.A. Kuratti and W. H. Sanders. Performance analysis of the RAID 5 disk array. Proc. lntl Comp. Performance and Dependability Syrnp., pages 236--45, Apr. 1995. Google ScholarDigital Library
- 20.S. Lavenberg, editor. Computer performance modeling hand* book. Academic Press, 1983. Google ScholarDigital Library
- 21.A. Merchant and P. S. Yu. Analytic modeling of clustered RAID with mapping based on nearly random permutation. IEEE Trans. on Computers, 45(3):367-73, Mar. 1996. Google ScholarDigital Library
- 22.S. W. Ng. Improving disk performance via latency reduction. IEEE Trans. on Computers, 40(1):22-30, Jan. 1991. Google ScholarDigital Library
- 23.W. Oney. Queueing analysis of the scan policy for movinghead disks. J. of the ACM, 22(3):397-412, July i975. Google ScholarDigital Library
- 24.C. Ruemmler and J. Wilkes. UNIX disk access patterns. Proc. Winter 1993 USENIX (San Diego, CA), pages 405- 20, 25-29 Jan. 1993.Google Scholar
- 25.C. Ruemmler and J. Wilkes. An introduction to disk drive modeling. IEEE Computer, 27'(3):17-28, Mar. 1994. Google ScholarDigital Library
- 26.B. Seeger. An analysis of schedules for performing multi-page requests. Information Sys., 21(5):387-407, July 1996. Google ScholarDigital Library
- 27.M. Seltzer, P. Chen, and J. Ousterhout. Disk scheduling revisited. Proc. Winter 1990 USENIX Conf. (Washington, DC), pages 313-23, 22----26 J#m. 1990.Google Scholar
- 28.E. Shriver. Performance modeling for realistic storage devices. PhD thesis. Dept of Comp. Science, New York University, May 1997. Google ScholarDigital Library
- 29..I.A. Solworth and C. U. Orji. Write-only disk caches. Proc. A CM SIGMOD Conf. (Atlantic City, NJ), pages 123-32, May t990. Google ScholarDigital Library
- 30.T. J. Teorey and T. B. Pinkerton. A comparative analysis of disk scheduling policies. Communications of the ACM, 15(3):177--84, Mar. 1972. Google ScholarDigital Library
- 31.F. Raab, editor. TPC benchmark C, Standard Specification Revision 3.0. Technical report. Transaction Processing Performance Council, 15 Feb. 1995.Google Scholar
- 32.N. C. Wilhelm. An anomaly in disk scheduling: a comparison of FCFS and SSTF seek scheduling using an empirical model for disk accesses. Communications of the A CM, 19(1):13-17, Jan. 1976. Google ScholarDigital Library
- 33.N. C. Wilhetm. A general model for the performance of disk systems. J. of the ACM, 24(1):14-31, Jan. 1977. Google ScholarDigital Library
- 34.J. Wilkes. The Pantheon storage-system simulator. Technical Report HPL-SSP-95-14. Storage Systems Program, Hewlett-Packard Labs., Pato Alto, CA, 29 Dec. 1995.Google Scholar
- 35.B. L. Worthington. Aggressive centralized and distributed scheduling of disk requests. PhD thesis, published as Technical report CSE-TR-244-95. Dept of Comp. Science and Eng., Univ. of Michigan, June 1995. Google ScholarDigital Library
- 36.B. L. Worthington, G. R. Ganger, and Y. N. Patt. Scheduling algorithms for modern disk drives. Proc. of A CM SIGMET- R ICS Conf. on Measurement and Modeling of Comp. Sys. (Nashville, TN), pages 241-251, May 1994. Google ScholarDigital Library
Index Terms
- An analytic behavior model for disk drives with readahead caches and request reordering
Recommendations
An analytic behavior model for disk drives with readahead caches and request reordering
Modern disk drives read-ahead data and reorder incoming requests in a workload-dependent fashion. This improves their performance, but makes simple analytical models of them inadequate for performance prediction, capacity planning, workload balancing, ...
Extending SSD lifetimes with disk-based write caches
FAST'10: Proceedings of the 8th USENIX conference on File and storage technologiesWe present Griffin, a hybrid storage device that uses a hard disk drive (HDD) as a write cache for a Solid State Device (SSD). Griffin is motivated by two observations: First, HDDs can match the sequential write bandwidth of mid-range SSDs. Second, both ...
DCD—disk caching disk: a new approach for boosting I/O performance
ISCA '96: Proceedings of the 23rd annual international symposium on Computer architectureThis paper presents a novel disk storage architecture called DCD, Disk Caching Disk, for the purpose of optimizing I/O performance. The main idea of the DCD is to use a small log disk, referred to as cache-disk, as a secondary disk cache to optimize ...
Comments