G. A. Alvarez, W. A. Burkhard, and F. Cristian. Tolerating multiple failures in RAID architectures with optimal storage and uniform declustering. In Proceedings of the 24th International Symposium on Computer Architecture, pages 62--72, Denver, CO, June 1997. ACM.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....Section 4.2.2. Two OBSDs are related if they store objects that are copies of each other or are part of the same redundancy set. We assume that two OBSDs are related at most once by making use of optimal data allocation algorithms [9] to ensure that the storage system is optimally declustered [2]. This distributes the work of reconstructing the objects on a failed OBSD optimally through the system. To speed up this reconstruction, we employ Fast Mirroring Copy (FMC) where the reconstructed objects are stored on different OBSDs throughout the system. This will repair an OBSD failure in ....

....scheme. This can be done by mirroring or triplicating data, using RAID Level 5 parity, or using erasure correcting codes such as Even Odd (which takes n blocks and adds to them two blocks such that any n survivors among the n blocks suffice to reconstruct all n 2) or Reed Solomon block codes [2, 19, 21]. This redundancy must be on different disks from the data it protects to guard against disk failure, but it might be possible to keep it on the same disk if it only must guard against nonrecoverable errors. 3.2. Galois Power Signatures of Storage Objects Storage objects are made up of blocks, ....

G. A. Alvarez, W. A. Burkhard, and F. Cristian. Tolerating multiple failures in RAID architectures with optimal storage and uniform declustering. In Proceedings of the 24th International Symposium on Computer Architecture, pages 62--72, Denver, CO, June 2002. ACM.

....the file grows [LMR98] Another avenue yet to explore is to keep each record in a single group, but to expand the parity schema to tolerate multiple failures within the group. Again the scalability requirement precludes a trivial application of earlier well known proposals, e.g. in [H al94] or [ABC97]. At present, ReedSalomon codes, already in commercial use for data storage systems where scalability is not yet a concern, I97] appear the best candidate. One potential hurdle, when scalability is taken into the account, is the efficiency of related operations in a tower of Galois fields ....

Alvarez, G., Burkhard, W., Cristian, F. Tolerating Multiple-Failures in RAID Architecture with Optimal Storage and Uniform Declustering. Intl. Symp. On Comp. Arch., ISCA-97, 1997.

....proposals for cases of correlated disk failures that may generalize to LH sa bucket failures. In [T98] there is an overview of several RAID oriented schemes seeking for 2 availability. Recently, the DATUM RAID scheme provides for n availability in an optimal amount of redundant storage [ABC97]. All these schemes address the static n. The already mentioned LH m , as well as LH s and LH g schemes, L a97] LR97] were first to address the high availability requirement in the context of horizontal scalability. One basic difference to the traditional and distributed RAID schemes is a ....

Alvarez, G., Burkhard, W., Cristian, F. Tolerating Multiple-Failures in RAID Architecture with Optimal Storage and Uniform Declustering. Intl. Symp. On Comp. Arch., ISCA-97, 1997.

....work Countless high availability schemes for centralized environments are known, using mirroring or striping, PGK88] HO95] T95a] The use of a mirror leads to k times higher storage overhead than for LH g. Unlike grouping, striping schemes basically deteriorate the search performance, W96] [ABC97]. Insert performance is also affected. No other scheme using the concept of record grouping is known. When a node failure is detected, some existing schemes redistribute data over available nodes, HST91] T95a] Others replace the unavailable node with a spare node where the unavailable ....

Alvarez, G., Burkhard, W., Cristian, F. Tolerating Multiple-Failures in RAID Architecture with Optimal Storage and Uniform Declustering. Intl. Symp. On Comp. Arch., ISCA-97, 1997.

....one error, even with periodic daily or weekly repairs, has a lower reliability than an individual disk. Their analysis assumes that erasures are detected when they occur; the truth for disk arrays is worsened if there are latent sector faults, which may remain undetected for long periods; see [1]. Most currently available systems handle only one disk failure [16] As arrays grow in size, the need for greater redundancy without a reduction in performance becomes apparent. Reliability is evidently a major concern. However, the tradeoff between reliability and performance has typically been ....

Guillermo A. Alvarez, Walter A. Burkhard and Flaviu Cristian. Tolerating multiple failures in RAID architectures with optimal storage and uniform declustering. Proceedings of the 24th Annual ACM/IEEE International Symposium on Computer Architecture, 1997.

....performance. The bucket recovery calculus can be made parallel between the participated buckets. The algorithms recovering from specific catastrophic failures can be added. Finally, the Reed Solomon Codes used are not the only possibility. Some other codes are potentially attractive as well, [ABC97], H a94] BFT98] The storage of parity buckets allows for interesting optimizations. The basic scheme stores parity buckets at a dedicated servers. This overhead to the number of servers may itself bother an application. Next, while the searches in normal mode do not concern the parity ....

....high availability at no additional hardware cost. 7 RELATED WORK There were countless high availability schemes for a single site, usually 1 available and using some RAID like striping. A few schemes appeared for the (static) k 1 k availability in this context, BM93] BBM93] H a94] and [ABC97] recently. There were also studies for the distributed environment, e.g. SG90] showing the inefficiency of any trivial striping. Deeper discussion of all these schemes, including SDDS schemes with mirroring or replication mentioned in the Introduction, is in [LMR98] However, besides the LH RS , ....

Alvarez, G., Burkhard, W., Cristian, F. Tolerating Multiple-Failures in RAID Architecture with Optimal Storage and Uniform Declustering. Intl. Symp. On Comp. Arch., ISCA-97, 1997.

....to nd an evenly distributed parity groups lookup table. Ng and Mattson [15] also proposed the same method in 1992. However, a balanced block can be too large and balanced incomplete block designs are not known for a large number of possible values of n and k. Alvarez, Burkhard, and Cristian[3] in 1997 presented the rst method that can mask any given number of failures. Their method requires an optimal amount of redundant storage space and spreads reconstruction accesses uniformly over all the disks in the presence of a disk failures without requiring large layout tables in the disk ....

..... Section 4 generalizes this to the case of n p m , for any integer m 2. Section 5 discusses the properties of the Prime groups. Section 6 concludes the paper. For most of this paper, we make the assumption below and the same assumption is made in all previous work that we know, including [3, 13, 9, 10]. A reason behind this assumption is that contiguous user data is typically distributed over the disks to improve the performance and this will make data accesses more evenly distributed. Assumption: Each disk has the same read and write load, and the amount of read and or write operations in a ....

[Article contains additional citation context not shown here]

G.A. Alvarez, W. A. Burkhard, F. Cristian, Tolerating Multiple Failures in RAID Architectures with Optimal Storage and Uniform Declustering, in Proceedings of the 24th Annual ACM/IEEE International Symposium on Computer Architecture (ISCA'97), Colorado, June 1997, pages 62-72.

....Level 5 [11] a stripe contains stripe units on all the disks. If a disk fails, all the other disks must process additional accesses for every read from the failed disk. The surviving disks then carry not only their own load, but also the read load of the failed disk. In a declustered disk array [1, 2, 6, 7, 10], the number of stripe units per stripe is smaller than the total number of disks. Any usable data layout (the assignment of stripe units to stripes) fulfills a number of fairly natural criteria: 1) Each stripe contains the same number k of stripe units. 2) Each disk contains the same number b ....

....units, and the actual placement of these stripe units on the various disks in the disk array. PDDL uses almost no storage or calculation to determine its mapping. Recent declustering schemes PRIME of Alvarez, Burkhard, Stockmeyer and Cristian [2] and DATUM of Alvarez, Burkhard, and Cristian [1] also use straightforward, efficient calculation to map stripe units to disks thereby avoiding table lookup. Previous schemes include Parity Declustering of Holland and Gibson [6] which uses table lookup to specify balanced incomplete block design together with parity rotation. The Random ....

G.A. Alvarez, W.A. Burkhard, F. Cristian: "Tolerating Multiple Failures in RAID Architectures with Optimal Storage and Uniform Declustering", Proceedings of the 24th Annual ACM/IEEE International Symposium on Computer Architecture, pp. 62-72, 1997.

....NVRAM cache in the order of tens of gigabytes. User data is ultimately stored in the back end disk drives, for a total capacity of many terabytes. For fault tolerance, arrays keep redundant data at both levels in the memory hierarchy: as mirror copies or erasure correcting codes on disks (RAID) [1, 16], and as dirty blocks mirrored in separate NVRAM cache banks in independent power domains. Disk arrays organize data storage into Logical Units (LUs) exporting a linear address space of blocks to client hosts. The NVRAM cache in the disk arrays serves several purposes. First, it acts as a ....

....storage array architectures. The physical characteristics and performance of MEMS based storage devices are discussed in several papers from the CMU Parallel Data Laboratory [3, 20, 8] The use of redundant data layouts for reliability, load balance and improved performance is well established [1, 2, 16], and these are commonly used in modern disk arrays. In most such layouts, the performance of the disk is limited by the disk head seek time and rotational delays, particularly for workloads with small, nonsequential I Os. Several mechanisms have been proposed to ameliorate the impact of ....

G.A. Alvarez, W.A. Burkhard, and F. Cristian. Tolerating multiple failures in RAID architectures with optimal storage and uniform declustering. In Proceedings of the 24th International Symposium on Computer Architecture (ISCA), pages 62--72. ACM Press, June 1997.

....units of an unavailable disk can be reconstructed via the redundancy of the scheme and stored in the allocated spare space. In a declustered data organization (originally suggested by Muntz and Liu [13] evaluated by Holland and Gibson [9, 10] and more recently improved upon by Alvarez et al.[1, 2], Merchant and Yu [12] Ng and Mattson [14] Reddy and Bannerjee [16] and Schwabe and Sutherland [17] each stripe is mapped to k of the n disks in the array (where k n) to achieve significant performance improvements during both degraded operation and on line disk reconstruction. These ....

....database.tar.Z. Merchant and Yu [12] propose the Pseudo Random scheme to replace the table lookup with on demand calculation using pseudo random permutations to obtain the mapping. Both parity and reconstruction workload are expected to be evenly distributed. Alvarez, Burkhard and Cristian [1] investigate the DATUM data layout which utilizes the binomial number system (based upon complete block designs) to obtain the mapping; the scheme calculates addresses on demand. Alvarez, Burkhard, Stockmeyer, and Cristian [2] study the PRIME and RELPR layouts which utilize on demand calculation ....

[Article contains additional citation context not shown here]

G.A. Alvarez, W.A. Burkhard, F. Cristian: "Tolerating Multiple Failures in RAID Architectures with Optimal Storage and Uniform Declustering", Proceedings of the 24th Annual ACM/IEEE International Symposium on Computer Architecture, p.62-72, June 1997.

....application of balanced incomplete block designs BIBD [11] to transaction processing with workload characteristics vastly different from those of video streams. This includes the work of Holland and Gibson [14] Ng and Mattson [18] Reddy, Chandy and Banerjee [22] Alvarez, Burkhard and Cristian [1], Alvarez, Burkhard, Stockmeyer, and Cristian [2] as well as Schwarz, Steinberg, and Burkhard [25] Disk array declustering for video servers based on BIBD has been considered by Cohen [9] and independently by Ozden, Rastogi, Shenoy and Silberschatz [20] The very recent overlay striping data ....

Alvarez, G.A., W.A. Burkhard, and F. Cristian, Tolerating Multiple Failures in RAID Architectures with Optimal Storage and Uniform Declustering. Proceedings of the 24th Annual International Symposium on Computer Architecture, pp. 62-72, 1997.

....and enduser capabilities, token economy, support for agents, and programmable network, will allow us to experimentally determine the effectiveness of these different strategies. Walt Burkhard is also investigating large scale data layout for multimedia applications, focusing on reliability issues [7] [38] In addition to experimenting with its storage facilities, the UCSD Active Web will provide a common testbed for both of these approaches to distributed multimedia storage research, supporting a large number of (possibly competing) storage intensive applications that can be observed and used ....

G.A. Alvarez, W.A. Burkhard, F. Cristian, "Tolerating Multiple Failures in RAID Architectures with Optimal Storage and Uniform Declustering," In Proc. 24th Annual Intl. Symp. on Computer Architecture, pages 62-72, 1997.

....and Sutherland [16] introduced BIBD based layouts for a restricted Technical Report CS98 576, CSE, UCSD, February 1998. 1 subset of the possible configurations, as well as perturbations that yield layouts for other cases with a bounded degradation in performance. Alvarez, Burkhard, and Cristian [2] presented layouts that compute on the fly mappings based on complete block designs, instead of storing block design tables in memory. There are six desirable properties of declustered disk array layouts that can tolerate a single failure, originally identified in [11] 1. Single failure ....

....as before, and the reconstruction workload is not perfectly uniform for nonprime n. We provide a bound on the reconstruction workload for general n. Independent of these considerations, there are numerous reasons for considering architectures capable of tolerating multiple concurrent failures [2, 3, 4, 7]. For example, there may be latent sector errors [19] that are discovered when the array already has suffered a failure and thus cannot tolerate another one, or communication failures that render disks inaccessible in network striping approaches [8] The PRIME and RELPR layouts can tolerate ....

[Article contains additional citation context not shown here]

G. Alvarez, W. Burkhard, and F. Cristian. Tolerating multiple failures in RAID architectures with optimal storage and uniform declustering. In Proc. of the International Symposium on Computer Architecture, pages 62--72, 1997.

No context found.

G. A. Alvarez, W. A. Burkhard, and F. Cristian. Tolerating multiple failures in RAID architectures with optimal storage and uniform declustering. In Proceedings of the 24th International Symposium on Computer Architecture, pages 62--72, Denver, CO, June 1997. ACM.

No context found.

G. A. Alvarez, W. A. Burkhard, and F. Cristian. Tolerating multiple failures in RAID architectures with optimal storage and uniform declustering. In Proceedings of the 24th International Symposium on Computer Architecture, pages 62--72, Denver, CO, June 2002. ACM.

No context found.

G. A. Alvarez, W. A. Burkhard, and F. Cristian. Tolerating multiple failures in RAID architectures with optimal storage and uniform declustering. In Proceedings of the 24th International Symposium on Computer Architecture, pages 62--72, Denver, CO, June 2002. ACM.

No context found.

G. A. Alvarez, W. A. Burkhard, and F. Cristian. Tolerating multiple failures in RAID architectures with optimal storage and uniform declustering. In Proceedings of the 24th International Symposium on Computer Architecture, pages 62--72, Denver, CO, June 2002. ACM.

No context found.

G. A. Alvarez, W. A. Burkhard, and F. Cristian. Tolerating multiple failures in RAID architectures with optimal storage and uniform declustering. In Proceedings of the 24th International Symposium on Computer Architecture, pages 62--72, Denver, CO, June 1997. ACM.

No context found.

G. A. Alvarez, W. A. Burkhard, and F. Cristian. Tolerating multiple failures in RAID architectures with optimal storage and uniform declustering. In ISCA '97, pages 62--72, 1997.

No context found.

G. A. Alvarez, W. A. Burkhard, and F. Cristian. Tolerating multiple failures in RAID architectures with optimal storage and uniform declustering. In ISCA '97, pages 62--72, 1997.

No context found.

G. A. Alvarez, W. A. Burkhard, and F. Cristian, "Tolerating Multiple Failures in RAID Architectures with Optimal Storage and Uniform Declustering," in Proceedings of the 24th ACM/IEEE International Symposium on Computer Architecture, pp. 62--71 (1997).

No context found.

G. A. Alvarez, W. A. Burkhard, and F. Cristian. Tolerating multiple failures in RAID architectures with optimal storage and uniform declustering. In Proceedings of the 24th International Symposium on Computer Architecture, pages 62--72, Denver, CO, June 1997. ACM.

No context found.

Guillermo A. Alvarez, Walter A. Burkhard, and Flaviu Cristian. Tolerating multiple failures in RAID architectures with optimal storage and uniform declustering. In Proceedings of the 24th Annual International Symposium on Computer Architecture (ISCA-97), volume 25,2 of Computer Architecture News, pages 62-72, New York, June 2-4 1997. ACM Press.

No context found.

G. A. Alvarez, W. A. Burkhard, and F. Cristian. Tolerating multiple failures in RAID architectures with optimal storage and uniform declustering. In Proceedings of the 24th Annual International Symposium on Computer Architecture, pages 62--72. IEEE Computer Society Press, 1997.

No context found.

G. A. Alvarez, W. A. Burkhard, and F. Cristian. Tolerating multiple failures in RAID architectures with optimal storage and uniform declustering. In Proceedings of the 24th Annual International Symposium on Computer Architecture, pages 62--72, Denver, CO, June 1997. IEEE Computer Society Press.

First 50 documents

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