Martin Schulze, Garth Gibson, Randy Katz, and David Patterson. How reliable is a RAID? Spring COMPCON'89 (San Francisco), pages 118--23. IEEE, March 1989.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....increased demand for better fault tolerance. Increasing the number of disks on which a file is stored in order to increase potential parallelism also increases the probability that the file will fail. The more devices a file is striped across, the higher the probability that one of them will fail [Schulze89]. Although the failure rates of secondary storage are decreasing dramatically [Gray90] relying on the reliability of individual devices is not a cost effective solution to the fault tolerance problem of disk arrays; rather, redundancy is needed [Patterson88] 7 1.1.4 The increasing demand for ....

Martin Schulze, Garth Gibson, Randy Katz, and David Patterson. How reliable is a RAID? Spring COMPCON'89 (San Francisco), pages 118--23. IEEE, March 1989.

....logging parity updates, we have reduced the disk time consumed by parity update by about a factor of eight. 2 1. Our failure model treats disk and controller failures as independent. If concurrent controller and disk failures must be survived, controller state must be partitioned and replicated [Schulze89, Gibson92, Cao93]. 2. Notice that we make no attempt to reduce the cost of the preread and overwrite of the target data block. Additional savings are possible if data writes and deferred and optimally scheduled [Solworth90, Orji93] D T V D TVD TVD TV 3V D T TVD TV D 10 ( 3V T 2 D 10 ( TVD 4 = TVD ....

Schulze, M. E., Gibson. G. A., Katz, R. H., and Patterson, D. A. How Reliable is a RAID? In Proceedings of the 1989 IEEE Computer Society International Conference (COMPCON 89) (1989). IEEE, 118-123.

....MTTF was the mean time to failure given that the upcoming failure is the one satisfying the condition. 131 An example of the cumulative conditional MTTF is mean time to critical (hazardous, unsafe) failure. Consider a system which has non critical failure states and a critical failure state [63, 98]. In [98] the authors classified failures in a RAID (Redundant Arrays of Inexpensive Disks) system. The recoverable system failure is due to recoverable data errors, and the catastrophic failure is due to unrecoverable data errors or drive failure. In [63] the safe failure state and the unsafe ....

....the mean time to failure given that the upcoming failure is the one satisfying the condition. 131 An example of the cumulative conditional MTTF is mean time to critical (hazardous, unsafe) failure. Consider a system which has non critical failure states and a critical failure state [63, 98] In [98], the authors classified failures in a RAID (Redundant Arrays of Inexpensive Disks) system. The recoverable system failure is due to recoverable data errors, and the catastrophic failure is due to unrecoverable data errors or drive failure. In [63] the safe failure state and the unsafe failure ....

M. Schulze, G. Gibson, R. Katz, and D. Patterson. How reliable is a RAID? In Proceedings of Spring COMPCON 89, pages 118--123, San Francisco, U.S.A., Mar. 1989.

....these disks. The ideas of disk interleaving and disk striping were first introduced by Kim [16] and Salem et al. [41] respectively. Since then, a great deal of work has focused on various design issues related to the performance of disk arrays [30, 29, 17, 20, 36, 4, 45] and to their reliability [43, 8, 27]. Disk array architectures fall into one of five different classes proposed in [35, 34, 14] referred to as Redundant Arrays of Inexpensive Disks (RAID) Among the five, the two most promising candidates for high performance computing systems appear to be the mirrored disk array (RAID 1) and the ....

Schulze, M., Gibson, G., Katz, R., and Patterson, D. How reliable is a RAID? In Proc. IEEE Spring ComCon Conf., pages 118--123, San Francisco, CA, February 1989.

....an exclusive or operation. have only 1 100th the reliability of a single disk [19] In large storage arrays, potential failures include transient media errors, media wear, head failure, other mechanical problems with the device, breakdown of disk controllers, and failed power supplies or cables [94]. There is error correction information embedded in each disk track that enables the disk to volunteer the information that it has an error. To correct disk errors, the disk array maintains additional parallel error correction information across the disks. Although it is not necessary to perform ....

Martin Schulze, Garth Gibson, Randy H. Katz, and David A. Patterson. How reliable is a RAID? In Proceedings IEEE COMPCON, pages 118--123, Spring 1989.

....there is simply no distinction between data and parity. It is this feature that makes it possible for IDA to be used not only to boost communication fault tolerance, but also to improve bandwidth allocation and utilization. An important aspect of IDA, unlike other redundancy injecting protocols [21, 16], is that the amount of redundancy to be used with a given object, or in a given communication session, does not have to be constant. In particular, our AIDA based bandwidth allocation strategy controls the amount of redundancy to be used with a particular object in a particular communication ....

Martin Schulze, Garth Gibson, Randy Katz, and David Patterson. How reliable is a RAID? In Proceedings of COMPCON-89, the Thirty-fourth IEEE Computer Society International Conference, March 1989.

....one of our contentions is that disks are no longer the primary cause of problems in a disk array. The reliability of the support hardware and the 1996 USENIX Technical Conference January 22 26, 1996 San Diego, CA array controller is little or no better than that of the disks. The data in [Schulze89] suggests that these support components would together lead to a mean time to failure of a small array of about 46k hours; Gibson93] simply increases this to a more reasonable value of 150k hours without further discussion. Fortunately, more recent designs and pressure by manufacturers to boost ....

Martin Schulze, Garth Gibson, Randy Katz, and David Patterson. How reliable is a RAID? Spring COMPCON'89 (San Francisco), pages 118--23. IEEE, March 1989.

....by MOSIS and tested in the VLSI lab of Harvard University, Cambridge, MA. The performance of the chip was measured to be about 1 megabyte per second. By using proper pipelining and more elaborate designs, this figure can be boosted significantly. Unlike other redundancy injecting protocols [23, 16], the amount of redundancy that IDA uses with a given object, or in a given session, does not have to be constant. In particular, as we will describe later, our AIDA based bandwidth allocation strategy controls the amount of redundancy to be used with a particular object in a particular session so ....

Martin Schulze, Garth Gibson, Randy Katz, and David Patterson. How reliable is a RAID? In Proceedings of COMPCON-89, the Thirty-fourth IEEE Computer Society International Conference, March 1989.

....paths, power supplies, etc. For the disk array to be fault tolerant it must be able to tolerate failures in any of these components. There are several known architectures for building disk array subsystems that are tolerant to failures in support hardware components and the disk drives [3, 16, 6, 9, 23]. The major support components of interest are disk controllers, cable linking a disk to the controller, power supply, etc. Currently, the existing architectures focus on improving the performance and or reliability of just the disk array or the supporting hardware. Some of the currently known ....

.... the power failure issues as they are easily addressed by using an uninterrupted power supply (UPS) Among the existing architectures that tolerate the failure of support hardware, the most prominent ones are Single Path Horizontal Array [6] Dual Path Vertical array [9] Dual Path Horizontal Array [23] and Crosshatch Disk Array [16] 2.2.1 Single Path Horizontal Array . Parity Group . C C C C C . Figure 3: Single path horizontal array. C C C Parity Group C C C . C C . C C ....

[Article contains additional citation context not shown here]

M. Schulze, G. A. Gibson, R. Katz, and D. A. Patterson, "How Reliable is a RAID," COMPCON 89, pp. 118-123, Feb-Mar 1989.

....redundancy is added uniformly; there is simply no distinction between data and parity. It is this feature that makes it possible for IDA to be used not only to boost communication fault tolerance, but also to improve bandwidth allocation and utilization. Unlike other redundancy injecting protocols [23, 16], the amount of redundancy that IDA uses with a given object, or in a given session, does not have to be constant. 4 For a concrete implementation and for examples, the reader is referred to our previous work on SETH [4] and IDA based RAID I O systems [5] 5 The chip (called SETH) has been ....

Martin Schulze, Garth Gibson, Randy Katz, and David Patterson. How reliable is a RAID? In Proceedings of COMPCON-89, the Thirty-fourth IEEE Computer Society International Conference, March 1989.

....7: Independence of Disk Failure e.g. a power supply, would result in the inaccessibility of an entire string of disks. Thus, disks sharing the same support hardware should not belong to the same disk array. In fact, the disk arrays should be constructed orthogonally to the support hardware groups [32]. In [14] the authors compare the MTTDL of an array with an non orthogonal organization to that of an array with an orthogonal organization and show a significant improvement in reliability. In addition to guarding against multiple failures due to a single support hardware failure, we would also ....

M. Schulze, G. Gibson, R. Katz, and D. Patterson. How-Reliable is a RAID? COMPCON, pages 118--123, 1989.

....11 October 28, 1993 MTTDL results. Since the need for large ensembles of disk arrays together [1] with the attractive prices for disks, obtaining a 10,000 year MTTDL is not impossible. Nevertheless approximately 8 of the disk arrays will experience triple failure during a MTTDL interval [7]; we have not totally solved the problem. However our work represents a step forward. To achieve these values, we can (1) provide spare strings, 2) shorten repair intervals, 3) harden disk drives, 4) harden strings or (5) increase data redundancy. We summarize the benefits and drawbacks with ....

Martin Schulze, Garth A. Gibson, Randy H. Katz, and David A. Patterson. How Reliable is a RAID? In Proceedings of the COMPCON Conference, pages 118--123, San Francisco, 1989.

....error correction group are inaccessible. Option 2, on the other hand, loses one disk from each of the four error correction groups and still allows access to all data. This technique of organizing error correction groups orthogonally to common hardware (such as a string) is called orthogonal RAID [Schulze89]. Orthogonal RAID has the added benefit of minimizing string conflicts when multiple disks from a group transfer data simultaneously. 4 ADVANCED TOPICS This section discusses advanced topics in the design of redundant disk arrays. Many of the techniques are independent of each other, allowing ....

Martin Schulze, Garth Gibson, Randy Katz, and David Patterson. How Reliable is a RAID? In Procedures of the IEEE Computer Society International Conference (COMPCON), March 1989. Spring COMPCON 89. Gives a reliability calculation for the electronics as well as the disks for RAIDs.

....error correction group are inaccessible. Option 2, on the other hand, loses one disk from each of the four error correction groups and still allows access to all data. This technique of organizing error correction groups orthogonally to common hardware (such as a string) is called orthogonal RAID [Schulze89, Ng94]. Orthogonal RAID has the added benefit of minimizing string conflicts when multiple disks from a group transfer data simultaneously. 4 ADVANCED TOPICS This section discusses advanced topics in the design of redundant disk arrays. Many of the techniques are independent of each other, allowing ....

Martin Schulze, Garth Gibson, Randy Katz, and David Patterson. How Reliable is a RAID? In Procedures of the IEEE Computer Society International Conference (COMPCON), March 1989. Spring COMPCON 89. Gives a reliability calculation for the electronics as well as the disks for RAIDs.

....workstation. configured as a RAID Level 5 [13] with one parity group of 24 disks. This scheme delivers high bandwidth but exposes the array to data loss during dependent failure modes such as a SCSI controller failure. Techniques for maximizing reliability are beyond the scope of this paper [4] [16], 6] For reads, data are read from the disk array into the memory on the XBUS board; from there, data are sent over HIPPI, back to the XBUS board, and into XBUS memory. For writes, data originate in XBUS memory, are sent over the HIPPI and then back to the XBUS board to XBUS memory; parity is ....

Martin Schulze, Garth Gibson, Randy H. Katz, and David A. Patterson. How Reliable is a RAID? In Proceedings IEEE COMPCON, pages 118--123, Spring 1989.

....Each region is a miniature replica of the array proposed above. Small user writes for a particular region are 2. Our failure model treats disk and controller failures as independent. If concurrent controller and disk failures must be survived, controller state must be partitioned and replicated [Schulze89, Gibson93, Cao93] 3. Notice that we make no attempt to reduce the cost of the overwrite of the target data block. Additional savings are possible if data writes are deferred and optimally scheduled [Solworth90, Orji93] D T V D TVD 2V V TVD TVD TV 3V D T TVD TVD TV D 10 ( 3V T 2 D 10 ( ....

Schulze, M. E., Gibson. G. A., Katz, R. H., and Patterson, D. A. How reliable is a RAID? Proceedings of the 1989 IEEE Computer Society International Conference (COMPCON 89) (Cat. No. 91TH0394-7), San Francisco CA, (Feb. 37-Mar. 3, 1989), IEEE Computer Society Press, Washington DC (1989) 118-123.

No context found.

Martin Schulze, Garth Gibson, Randy Katz, and David Patterson, How Reliable is a RAID?, COMPCON-89, the Thirty-fourth IEEE Computer Society International Conference, March 1989.

No context found.

Martin Schulze, Garth Gibson, Randy Katz, and David Patterson, How Reliable is a RAID?, COMPCON-89, the Thirty-fourth IEEE Computer Society International Conference, March 1989.

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