H. Simitci and D. A. Reed. Adaptive disk striping for parallel input/output. In Proceedings of the Seventh NASA Goddard Conference on Mass Storage Systems, pages 88--102, San Diego, CA, 1999. IEEE Computer Society Press. Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
Improving Disk Throughput in Data-Intensive Servers - Carrera, Bianchini (2003) (Correct)
....for balancing the load is striping. Striping groups several sequential disk blocks in units of fixed size and lays those units out across the physical disks in round robin fashion. Smaller striping units normally lead to better load balancing. However, very small units can hurt performance [28], as each large request gets fragmented into sub requests to multiple disks. In the absence of contention, the response time of a request for r blocks can be defined as T (r) D) T (r=D) where D is the number of sub requests and (D) is a factor that depends on the probability distribution ....
.... In the absence of contention, the response time of a request for r blocks can be defined as T (r) D) T (r=D) where D is the number of sub requests and (D) is a factor that depends on the probability distribution function of T (r=D) For instance, D) 2D= D 1) for a uniform distribution [28]. For the purposes of this paper, the most important aspect of striping is its relationship with disk read ahead. Since striping spreads files across multiple disks, it increases the chance that consecutive blocks of a disk do not belong to the same file. Thus, read aheads that are larger than ....
H. Simitci and D. A. Reed. Adaptive disk striping for parallel input/output. In Proceedings of the Seventh NASA Goddard Conference on Mass Storage Systems, pages 88--102, San Diego, CA, 1999. IEEE Computer Society Press.
Improving Disk Throughput in Data-Intensive Servers - Carrera, Bianchini (2002) (Correct)
....increases throughput up to a point. The main reason why increasing the striping unit improves throughput is that it allows larger disk reads, which in turn reduce the overall latency of the disk array by a greater factor than the faster parallel transfers. This effect has been recognized before in [25]. The improvement in throughput stops when the striping unit becomes larger than the average size of files. Increasing the multithreading degree improves throughput for a different reason: more threads allow for better disk scheduling. 0 40 80 120 160 LDF FOR Basic Segm Figure 9: VoD ....
Huseyin Simitci and Daniel A. Reed. Adaptive disk striping for parallel input/output. In Proceedings of the Seventh NASA Goddard Conference on Mass Storage Systems, pages 88--102, San Diego, CA, 1999. IEEE Computer Society Press.
Manageable Storage via Adaptation in WiND - Arpaci-Dusseau, Arpaci-Dusseau, .. Self-citation (Storage) (Correct)
....in mirrored storage for improved performance, and cold data is moved into RAID 5 storage to increase effective capacity. Such a performance versus capacity optimization could be placed into the GALE framework. Adaptivity has also been explored within the context of parallel file systems [27]. In that study, the authors explore the use of fuzzy logic to adaptively select the proper stripe size for a storage system. Such approaches may also be applicable within WiND. Long term adaptation has shown promise in single disk log structured file systems [20] With this approach, file layout ....
H. Simitci and D. A. Reed. Adaptive disk striping for parallel input/output. In Seventh NASA Goddard Conference on Mass Storage Systems, San Diego, CA, March 1999. IEEE Computer Society Press.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC