K. Salem and H. Garcia-Molina, "Disk striping," in Proceedings of the 2nd International Conference on Data Engineering, pp. 336--42, IEEE, 1986.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....we get a starting point for the design of a general purpose parallel disks server based on the scheduling algorithms presented here. Section 8 summarizes the results and mentions some open questions. 1. 3 Related Work An automatic load balancing approach widely used in practice is striping [3, 4]. In our terminology that means a logical block size D times larger than the physical block size, where each logical block is dispersed over all disks. This works for scanning large amounts of consecutive data but is of little help for smaller access granularity. Without redundancy, worst case ....

K. Salem and H. Garcia-Molina, "Disk striping," Proceedings of Data Engineering'86, pp. 336-- 342, 1986.

....evenly across multiple disks such that whenever that storage unit is requested, concurrent transfers of blocks from each of the disks occurs. However, average access times can be large in disk slriping systems since access time is governed by the disk with the largest access time. Garcia Molina [7] proposed a solution based on immediate reading which allows a disk read to occur anywhere within a block, and this was found to improve performance considerably. Kim [10] developed a solution to 56 this problem based on hardware synchronization of the disks. Disk striping has difficulty in ....

Garcia-Molina, H., and Salem, K., "Disk Striping.", International Conference on Data Engineering, February 1986, pp. 336-342.

....advantages over traditional secondary storage systems. First, the data on the disks can be accessed in parallel, which improves the I O performance. Each file that is stored in the array is striped and placed on several disks. This scheme improves the response time when the user accesses that file [Kim86, Reddy89, Salem86]. The controller can also carry out several operations at the same time if the disks involved in these operations are different. This scheme improves the throughput. Second, when the number of disks increases in a disk array, the availability of data and the reliability of the disk array may ....

K. Salem and H. Garcia-Molina. "Disk Striping". Proceedings of the 2nd International Conference on Data Engineering, 1986, pp. 336--342. 31

....multiple disks because it will never exercise more than one disk at a time. In contrast, prefetching on behalf of an application will increase the frequency with which there are multiple I O requests outstanding at the same time. If data is well distributed across multiple disks (where striping [47] generally suffices to produce such a distribution) it is likely that different requests can often be serviced in parallel by multiple disks. Therefore, prefetching increases the degree to which applications can exploit multiple disks. Conversely, as described above, if an application s ....

K. Salem and Hector Garcia-Molina. Disk striping. In Proceedings of the 2nd IEEE International Conference on Data Engineering, 1986.

....validated by previous performance studies [Corti,1990] A database processor (DP) is used to handle normal database processing, while a recovery processor (RP) is used to perform recovery processing activities. HOST DP RP Figure 1 MMDB Model The AM structure 1s designed using disk atriping [Salem, 1986]. Each disk is divided into two areas system data cylinders and user data cylinders The system data cylinders consist of data pertaining to system information, such as data dictionary, and address translation tables, while the user data cylinders contain the user data only. The system data ....

Kenneth Salem and Hector Garcta-Molina, "Disk Striping," Proceedings of the [EEE International Conference on Data Engineering, February 5-7, 1986, pp. 336-342

....1. An appropriate candidate for SS appears to be arrays of magnetic disks [PGK88] Given the high bandwidth requirements of many multimedia objects and the typical, relatively low, effective bandwidth of magnetic disks (nominal bandwidths are in the range of 4 15 MB sec) striping techniques [SGM86,BMGJ94,TF98] are likely to prove beneficial, since, when striping objects across several disks, the effective SS bandwidth for these objects is several tape CDROM based elevate disk based (elevate) Figure 1. Conceptual model of a Hierarchical Storage Management System (HSMS) times that ....

....to increase the available SS bandwidth in the system. One thread of this research led to the development of disk arrays [PGK88] and to the development of new methods for placing data objects on disk arrays, which attempt to exploit their inherent potential for high performance and reliability [SGM86,PGK88,CP90, Gea94] mainly through data placement techniques, such as striping. Multimedia data, such as video and audio, require very large storage capacities and very large I O bandwidths, making disk arrays the natural choice for the SS medium. Researchers in multimedia and video storage ....

Kenneth Salem and Hector Garcia-Molina. Disk striping. In Proc. International Conference on Data Engineering, Los Angeles, CA, pages 336--342, February 1986.

....generation. This document was written during Soraya Zertal s academic visit to the departement of computing, Imperial College of Science, Technology and Medicine, London SW7 2BZ. 1 1 Introduction Towards the end of the 80s, Redundant Arrays of Independent (or Inexpensive) Disks (RAID) [11, 6, 2, 3, 9] were proposed as a solution to the storage requirements problems. RAID provides a large storage capacity by grouping a large number of disks, a high I O performance by exploiting parallel access to disks, and a high data availability using integral or partial redundancy. The distributed and ....

K. Salem and G. Molina. Disk striping. In Proceedings of International Conference on Data Engineering, February 1986.

.... architects are able to increase the overall performance of the disk system by placing data across the drives in a manner that either enables independent accesses to perform positioning operations concurrently, or enables large accesses to transfer data from several drives concurrently [Kim86, Salem86] By striping data such that a unit of access spans all drives, the array offers an effective bandwidth equal to the transfer rate of a single disk multiplied by the number of disks in the array. Because all disks transfer data in parallel, this disk array architecture is known as a ....

Salem, K. and Garcia-Molina, H. "Disk striping." Proceedings of the 2nd International Conference on Data Engineering. Los Alamitos, CA: IEEE Computer Society Press. (1986) 336-342.

....think time in the system before being issued back into the subsystem model. For example, non zero think times might be used to represent the processing of one block before accessing the next. While rare, non zero think times have been used in published storage subsystem research. For example, [Salem86] uses a closed workload of read block, process block, write block (optional) sequences to emulate a generic file processing application. 17 spersed with long periods of idle time (i.e. no incoming requests) With a constant number of requests in the system, there is no burstiness. Workload ....

....drives, tends to suffer from substantial load balancing problems and does not allow multiple drives to cooperate in servicing requests for large amounts of data. Disk striping (or interleaving) spreads logically contiguous data across multiple disks by hashing on the logical address [Kim86, Salem86]. The performance impact of disk striping has been studied with both open subsystem models [Kim86, Kim91] and closed subsystem models [Salem86] The load balancing benefits of disk striping have been demonstrated with open subsystem models [Livny87, Ganger93a] The stripe unit size (i.e. the ....

[Article contains additional citation context not shown here]

K. Salem, G. Garcia-Molina, "Disk Striping", IEEE International Conference on Data Engineering, 1986, pp. 336--342.

....the contents of the dead disk onto the new one by reading the other N 1 data disks plus the parity disk. If a user read occurs before a disk block is reconstructed, then the above procedure must be followed to obtain the lost contents on the fly. The RAID controller will automatically stripe [LIVN86, SALE86] data across all N data disks on a track by track basis. The reason for track striping rather than sector striping is to obtain greater bandwidth 2 on sequential reads. With block striping, each disk will return one sector every rotation. With track striping, it can easily do track at a time ....

Salem, K. and Garcia-Molina, H., "Disk Striping," Proc. 1986 IEEE Data Engineering Conference, Los Angeles, CA, February 1986.

....on a single drive. If necessary, it would be broken into multiple smaller contiguous extents. In our storage system an extent of size E would correspond to a vertical rectangle of width 1 and height E. 10 Recently, researchers have suggested striping files across a collection of disks [SALE86, LIVN85]. Striping L = D disks entails allocating the Ith track to the Jth disk determined by: J = remainder (I L) 1 In this way, a large sequential I O can be processed by reading all disks in parallel, and very high bandwidth on sequential I O is possible. In our model this corresponds to a ....

Salem, K. and Garcia-Molina, H., "Disk Striping," Proc. 1986 IEEE Data Engineering Conference, Los Angeles, CA, February 1986.

....of the create call and specific fcntl features used for PVFS metadata and partition parameters. 2.2. Striping and Partitioning in PVFS PVFS provides two primary services over traditional file systems. First, it allows data in a file to be striped across multiple disk volumes on multiple nodes [4]. Striped files can be accessed with a UNIX style file system interface. Striping parameters are used to define how data is distributed to the I O nodes. Default striping parameters can be used or an fcntl( call can be used to specify custom striping parameters. Second, an application can ....

K. Salem and H. Garcia-Molina, "Disk Striping," IEEE 1986 Conference on Data Engineering, pp 336-342, 1986.

....storage subsystems, it is also necessary to increase their raw input output capacities and bandwidths. To support multi teraflop applications manipulating multi petabyte data sets, next generation file systems will have to stripe data over thousands of secondary and tertiary storage devices [4, 5]. Disk striping is such a tech1 Timeline 1998 2002 2006 Compute Performance 1.5 teraops 50 teraops 400 teraops Machine I O Rate 3 GB sec 100 GB sec 800 GB sec Typical Data Query 15 TB 500 TB 4 PB Total Transfer Time 5000 sec 5000 sec 5000 sec Table 1.1: Trends in computing and storage ....

....striping effective. The work on flexible parallel file systems and computational steering for adaptive control is presented in x2.2. 2. 1 Disk Striping One way to decrease the performance difference between computing elements and storage devices is to stripe data across several storage devices [5], and access them in parallel to achieve higher input output rates. Striping systems combine the storage space of multiple disks into a single large logical disk. Then, this logical disk space is spread across the physical disks in a round robin fashion, one stripe unit at a time. In chapter 1, ....

K. Salem and H. Garcia-Molina, "Disk Striping," in Proceedings of the 2 nd International Conference on Data Engineering, pp. 336--342, ACM, Feb. 1986.

....fragments at different network sites [RIES78] Data partitioning allows parallel database systems to exploit the I O bandwidth of multiple disks by reading and writing them in parallel. This approach provides I O bandwidth superior to RAID style systems without needing any specialized hardware [SALE84, PATT88]. The simplest partitioning strategy distributes tuples among the fragments in a roundrobin fashion. This is the partitioned version of the classic entry sequence file. Round robin partitioning is excellent if all applications want to access the relation by sequentially scanning all of it on ....

Salem, K. and H. Garcia-Molina, "Disk Striping", Department of Computer Science Princeton University Technical Report EEDS-TR-332-84, Princeton N.J., Dec. 1984

....and scalable solution. To store a large media object X on such a platform it is commonly striped into n blocks: X 0 , X 1 , X n 1 [Pol91] There are two basic techniques to assign these blocks to the magnetic disk drives that form the storage system: a) in a round robin sequence [SGM86], or (b) in a random manner [MSB97] Traditionally, the round robin placement utilizes a cycle based approach [GZS 97] to scheduling of resources to guarantee a continuous display, while the random placement utilizes a deadline driven approach [RW94] We have previously implemented paradigm (a) ....

Salem, K.; Garcia-Molina, H. "Disk Striping." Proceedings of International Conference on Database Engineering, February, 1986.

....benchmarks quickly become obsolete as machines evolve. For instance, IOStone tries to exercise the memory hierarchy but touches only 1 MB of user data. Perhaps at the time IOStone was written 1 MB was a lot of data but no longer. One recent example of how I O systems are evolving is disk arrays [Patterson88, Gibson91, Chen90, Salem86]. Disk arrays allow multiple I Os to be in progress simultaneously. Most current I O benchmark do not scale the number of processes issuing I O, and hence are unable to properly stress disk arrays. Unfortunately, it is difficult to find widespread agreement on a scaling strategy, especially for ....

K. Salem and H. Garcia-Molina, "Disk Striping", Proceedings of the Second International Conference on Data Engineering, 1986, 336-342.

....in integrating multimedia data into communications and computing has lead to a flurry of activity 161 in supporting high performance I O that satisfies special requirements of this data. Here we summarize some notable efforts. 7.9. 1 High Bandwidth Disk I O for Supercomputers Salem et al. [100] represents some of the early ideas on using disk arrays and associated data striping schemes to improve effective storage throughput. Observing that large disk arrays have poor reliability and that small disks outperform expensive highperformance disks in price vs. performance, Patterson et al. ....

Salem, K., and Garcia-Molina, H., "Disk Striping," IEEE International Conference on Data Engineering, 1986. 199

....power. Is it now very important to balance the I O bandwidth and the processing power in order to improve the overall performance of a system [6, 14] This research has been supported in part by KOSEF, ADD, IITA and ETRI, 1996. 1 There have been much research in improving I O performance [7, 8, 12, 14, 15], known as data declustering and disk striping in disk array systems. All of them utilizes the parallelism among many disks. However, incorporating such a large number of disks into a disk array system makes the disk array system more prone to failure than a single disk [12] High reliability can ....

Salem, Kenneth, and Hector Garcia-Molina, "Disk Striping," in Proc. Intl. Conf. Data Engineering, 1986.

No context found.

K. Salem and H. Garcia-Molina, "Disk striping," in Proceedings of the 2nd International Conference on Data Engineering, pp. 336--42, IEEE, 1986.

No context found.

K. Salem and H. Garcia-Molina, "Disk striping," in Proc. IEEE Data Engineering, pp. 336--342, Feb. 1986.

No context found.

Salem K. and Garcia-Molina H., "Disk Striping," Proceedings 1986 Data Engineering Conf, Los Angeles, February 1986.

No context found.

Salem, K. and H. Garcia-Molina, "Disk Striping", Department of Computer Science Princeton University Technical Report EEDS-TR-332-84, Princeton N.J., Dec. 1984

No context found.

K. Salem and H. Garcia-Molina, "Disk striping," in Proceeding of the IEEE 2nd International Conference on Data Engineering (ICDE'86), Los Angeles, CA, Feb. 1986, pp. 336--342.

No context found.

Kenneth Salem and Hector Garcia-Molina. Disk striping. In IEEE 1986.

No context found.

Salem, K., and H. Garcia-Molina, "Disk Striping", Technical Report No. 332, EECS Department, Princeton University, December 1984.

First 50 documents  Next 50

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