A. L. N. Reddy and P. Banerjee. Gracefully Degradable Disk Arrays. In FTCS-21, pages 401--408, Montreal, Canada, June 1991.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....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 layouts are based on balanced incomplete block designs (BIBDs) 7, 8] ....

....on demand calculation to obtain the mapping. These schemes come very close to meeting the layout goals that we discuss below. Schwabe and Sutherland [17] introduce a slightly relaxed BIBD designs which obtain approximately balanced layouts implemented via on demand calculation. Reddy and Banerjee [16] as well as Ng and Mattson [14] also present declustering schemes based upon BIBDs. All these schemes are logically correct; however, the run time performance varies dramatically depending upon the BIBD. Some of the earlier schemes do not depend upon particular families of BIBDs; the more recent ....

A.L.N. Reddy, P. Bannerjee: "Gracefully Degradable Disk Arrays", Proc. 21st International Symposium on Fault Tolerant Computing, pp. 501-506, April 1994.

....to a block on the failed disk causes one request to be sent to each of the surviving disks. Consequently, each surviving disk must service its own set of requests, an equivalent stream generated by requests on to the failed disk, and the requests generated by the background reconstruction process [18, 28]. The declustered parity disk array organization [17, 22, 24, 28] addresses this problem by trading some of the array s capacity for improved performance in the presence of disk failures. Specifically, it requires that each parity block protect some smaller number of data blocks (say (G Gamma ....

....to each of the surviving disks. Consequently, each surviving disk must service its own set of requests, an equivalent stream generated by requests on to the failed disk, and the requests generated by the background reconstruction process [18, 28] The declustered parity disk array organization [17, 22, 24, 28] addresses this problem by trading some of the array s capacity for improved performance in the presence of disk failures. Specifically, it requires that each parity block protect some smaller number of data blocks (say (G Gamma 1) By appropriately distributing the parity information across ....

A. L. Narasimha Reddy and P. Banerjee. Gracefully Degradable Disk Arrays. In Proceedings of the International Symposium on Fault Tolerant Computing, pages 401--408, June 1991.

....parity groups that appears in a row of the lookup table. The value of k 1 n 1 determines the fraction of each surviving disk that must be read during the reconstruction of a failed disk. This value is called the declustering ratio in [14] To achieve some of the layout goals, Reddy and Banerjee [17] proposed a method to solve the distribution problem. Their method uses some results of Hadamard matrices and di erence sets to solve the problem for n 2 . Merchant and Yu [12] proposed a method which uses Thorps s shu e to distribute the parity groups evenly. The method works for any n and k, ....

A.L. Reddy and P.Banerjee, Gracefully Degradable Disk Arrays. In Digests of Papers of the 21st International Symposium on Fault-Tolerant Computing, Pages 401-408, The Computer Society, IEEE, June, 1991.

....1.0) parity declustering is equivalent to RAID Level 5. Thus parity declustering can be seen as defining a continuum of design points between RAID Level 5 and mirroring, exhibiting improved performance in the presence of failure but increased capacity overhead as G is reduced. Reddy and Bannerjee [Reddy91] proposed an organization essentially equivalent to a 1 2 0 Logical Array Physical Array 0 1 2 3 4 5 6 S S S S S S S 3 S Figure 3.2: Declustering a parity stripe of size four over an array of seven disks. C G 1 4 1 51 limited version parity declustering. Their technique assigns each unit on ....

....incomplete block designs, and contrast it to a layout proposed by Merchant and Yu [Menon92a] which supports more combinations of array size (C) and number of units per parity stripe (G) in large arrays, at the cost of higher complexity. 3.2.1. Layout goodness criteria Extending from prior studies [Lee90, Dibble90, Reddy91, Merchant92a], we have identified six criteria for a good disk array layout. 1. Single failure correcting. No two units (whether data or parity) contained in the same parity stripe may reside on the same physical disk. This is the basic characteristic of any failure tolerating organization. In arrays in which ....

A.L.N. Reddy and P. Bannerjee, "Gracefully Degradable Disk Arrays," Proceedings of the International Symposium on Fault-Tolerant Computing, 1991, pp. 401408.

....parity updates on other nearby activity done in the foreground; batching together updates that are physically close together; or simply doing a single, linear sweep through the disks. Similarly, existing schemes for balancing disk traffic under failure conditions can be applied to AFRAID (e.g. [Gray90c, Muntz90, Blaum94, Reddy91]) For ease of exposition, however, we concentrate here on a straightforward left symmetric RAID 5 data layout. 3. Availability model of AFRAID In this section we develop analytic models of data loss mechanisms for AFRAID, basing them on similar models for traditional RAIDs. In the next section ....

A. L. Narasimha Reddy and P. Banerjee. Gracefully degradable disk arrays. Proceedings of FTCS-21, pages 401--8. Institute of Electrical and Electronics Engineers, June 1991.

....delays is an effect that is difficult to model analytically, but relatively straightforward to address in a simulation based study. In this paper we will use balanced incomplete and complete block designs (described in Section 4.2) to achieve better performance during reconstruction. Reddy [Reddy91] has also used block designs to improve the recovery mode performance of an array. His approach generates a layout with properties similar to ours, but is restricted to the case where G = C 2. 4. Data layout strategy In this section, we describe our layout goals, and the technique used to ....

A. Reddy and P. Bannerjee, "Gracefully Degradable Disk Arrays," Proceedings of FTCS-21, 1991, pp. 401-408.

....group length. In disk arrays, spare space is provided for reconstructing the failed data during the reconstruction process. This spare space can be organized in several ways. We consider three alternatives proposed in the literature: hot sparing [3] distributed sparing [5] and parity sparing [6]. In hot sparing, a hot spare disk is used to recover from failures. Hot sparing has some drawbacks: the failure of a spare disk may go unnoticed since it is not used during normal operation and hence can cause an unrecoverable loss if another disk fails after an undetected failure of the spare ....

....P0 S D0 D0 D0 D0 P0 D0 S D0 D0 D0 P0 D0 D0 S D0 D0 P0 D0 D0 D0 S D0 P0 D0 D0 D0 D0 S P0 D0 D0 D0 D0 D0 S D0 D0 D0 D0 D0 P0 D0 D0 D0 D0 P0 D0 D0 D0 D0 P0 D0 D0 D0 D0 P0 D0 D0 D0 D0 P0 D0 D0 D0 D0 P0 D0 D0 D0 D0 D0 Fig. 1. Array organization before and after a failure in hot sparing. Parity sparing [6] proposes an approach where the disk array has no spare disks. The spare disk is used to provide a second parity disk and thus to reduce the parity group length. On a failure, the parities of the two groups are merged to obtain a single larger array with a single parity group with a larger parity ....

[Article contains additional citation context not shown here]

A. L. Narasimha Reddy and P. Banerjee. Gracefully degradable disk arrays. Proc. of FTCS-21, June 1991.

No context found.

A. L. N. Reddy and P. Banerjee. Gracefully Degradable Disk Arrays. In FTCS-21, pages 401--408, Montreal, Canada, June 1991.

No context found.

A. L. N. Reddy and P. Banerjee. Gracefully Degradable Disk Arrays. In FTCS-21, pages 401--408, Montreal, Canada, June 1991.

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