RAID5 tolerates single disk failures by exclusive-ORing (XORing) the blocks corresponding to a requested block on the failed disk to reconstruct it. This results in increased loads on surviving disks and degraded disk response times with respect to normal mode operation. Provided a spare disk is available, a rebuild process systematically reads successive disk tracks, XORs them to recreate lost tracks and writes them onto a spare disk, thus returning the system to its original state. Rebuild time is important since RAID5 disk arrays with a single disk failure are susceptible to data loss if a second disk fails. According to the vacationing server model (VSM), rebuild read requests on surviving disks are given a lower priority than external user requests, so as to have less impact on their response time. Given that disk loads are balanced due to striping, rebuild time can be approximated by the time to read the contents of any one of the surviving disks. The analysis of the M/G/1 queueing model of VSM, given in this article, is more accurate and yet simpler than a previous analysis, but it also takes into account the effect of disk zoning explicitly. We also present a heuristic method to estimate rebuild time, which can be combined with the new analysis. The ability to quickly and accurately estimate rebuild time is useful in computing the reliability of RAID5 systems, especially during design tradeoff studies. The accuracy of the various analyses to estimate rebuild time are checked against detailed simulation results.

Redundant disk array architecture provides fault tolerance against disk drive failures. However, a storage subsystem consists of more than just disk drives. There must also be controllers for interfacing with the disk drives, cabling for providing data/control paths, power supplies, etc. For the subsystem to be fault tolerant, it must also be able to tolerate failure in any one of these components. While currently known array architectures can be designed to be tolerant to single failure in the support hardware, little attention has been paid to performance in the event of such failures. The recovery procedures required after such failures have been repaired were also not fully considered. The author presents a new array architecture which outperforms other currently known array architectures when some supporting hardware such as a controller or cable has failed. Furthermore, this architecture is fault tolerant to at least double failures in the support hardware and is, therefore, a more reliable and robust architecture.>

Several new disk arrays have recently been proposed in which the parity groupings are uniformly distributed throughout the array so that the extra workload created by a disk failure can be evenly shared by all the surviving disks, resulting in the best possible degraded mode performance. Many arrays now also put in multiple spare disks so that expensive service calls can be deferred. Furthermore, in a new sparing scheme called distributed sparing, the spare spaces are actually distributed throughout the array. This means after a rebuild the new array will be logically different from the original array. The authors present an algorithm for constructing and maintaining arrays with distributed sparing so that repeated uniform parity group distribution is achieved with each successive failure.>

Disk arrays are increasingly being used in distributed computing systems, as the vehicle for providing reliable and high performance data storage. When a disk in a RAID-5 fails, data in that failed disk can still be made available through parity reconstruction by reading from the other disks. However, this poses an increased burden on the surviving disks, and if consideration is not given to this failure consequence, then the performance of the system may degrade to an unacceptable level. This paper describes techniques that will enable the disk array to maintain good performance in the event of a disk failure. After a failed disk has been repaired, its content must be reconstructed from all the associated parity groups. In RAID-5, this must be a single thread sequential process. With the techniques introduced in this paper, it is shown how this sequential process can now be broken down into multiple parallel processes distributed throughout the array, thus shortening the reconstruction time. While the techniques introduced in this paper are applied to disk arrays, they may potentially have general applications in other areas of distributed computing.>

It is shown that in many environments the rotational latency and the rotational position sensing (RPS) miss delay are the major contributors to a disk's basic service time. A sensitivity study using a simple analytical queueing model shows that a reduction in these two components (both of which are related to the rotation of disk drives) has the greatest impact in reducing the disk's basic service time and in turn produces the greatest improvement in overall subsystem performance. While the most straightforward way to reduce latency and RPS miss penalty would be to increase the disk's rotation speed, there are some limitations to such an approach. Several alternatives to reducing latency and RPS miss penalty are proposed and explored, and their performance is analyzed using analytical queuing models.>

Four alternative implementations for achieving higher data rates in a disk subsystem (parallel heads without replication, parallel heads with replication, parallel actuators without replication, and parallel actuators with replication) are studied. Focus is on the tradeoffs between the number of devices and the number of data paths while keeping the number of physical devices constant (which may keep the cost roughly constant). The performance advantages and limitations of the alternative implementations are analyzed using an analytic queuing model and compared to a conventional disk subsystem. The study shows that parallel heads with replication from a single actuator performs the best for the average application environments, although other configurations may be more cost-effective.>

Two classes of simple low-rate burst-correcting convolutional codes that meet the Gallager bound ofg/b \geq 1 + R/1 - Rare presented. Special cases of these codes are also shown to meet the Peterson and Weldon bound ofb \leq [\frac{1}{2}(n_A -- 1)]. A third class of low-rate codes that can correct iow-density typeB2bursts is also presented. All three classes of codes are majority-logic decodable and, hence, can be easily implemented.

In many communication channels noise disturbances occur in the form of low-density bursts. Codes capable of correcting multiple bursts of low density are efficient tools of error control for such channels. One such class of codes is presented ia this paper. This Class of codes can be decoded by simple majority logic. The approach is to use codes which are the duals of product codes whose component codes are majority-logic-decodable cyclic codes.

In this correspondence, a modification of Rudolph's one-step majority-logic decoding algorithm is introduced. Using this modification, it is proved that all single-error-correcting codes can be one-step majority decoded.