Disk arrays were proposed in the 1980s as a way to use parallelism between multiple disks to improve aggregate I/O performance. Today they appear in the product lines of most major computer manufacturers. This paper gives a comprehensive overview of disk arrays and provides a framework in which to organize current and future work. The paper first introduces disk technology and reviews the driving forces that have popularized disk arrays: performance and reliability. It then discusses the two architectural techniques used in disk arrays: striping across multiple disks to improve performance and redundancy to improve reliability. Next, the paper describes seven disk array architectures, called RAID (Redundant Arrays of Inexpensive Disks) levels 0-6 and compares their performance, cost, and reliability. It goes on to discuss advanced research and implementation topics such as refining the basic RAID levels to improve performance and designing algorithms to maintain data consistency. Last, the paper describes six disk array prototypes or products and discusses future opportunities for research. The paper includes an annotated bibliography of disk array-related literature.

This paper describes the design of the Gamma database machine and the techniques employed in its implementation. Gamma is a relational database machine currently operating on an Intel iPSC/2 hypercube with 32 processors and 32 disk drives. Gamma employs three key technical ideas which enable the architecture to be scaled to 100s of processors. First, all relations are horizontally partitioned across multiple disk drives enabling relations to be scanned in parallel. Second, novel parallel algorithms based on hashing are used to implement the complex relational operators such as join and aggregate functions. Third, dataflow scheduling techniques are used to coordinate multioperator queries. By using these techniques it is possible to control the execution of very complex queries with minimal coordination- a necessity for configurations involving a very large number of processors. In addition to describing the design of the Gamma software, a thorough performance evaluation of the iPSC/2 hypercube version of Gamma is also presented. In addition to measuring the effect of relation size and indices on the response time for selection, join, aggregation, and update queries, we also analyze the performance of Gamma relative to the number of processors employed when the sizes of the input relations are kept constant (speedup) and when the sizes of the input relations are increased proportionally to the number of processors (scaleup). The speedup results obtained for both selection and join queries are linear; thus, doubling the number of processors

Multimedia information systems have emerged as an essential component of many application domains ranging from library information systems to entertainment technology. However, most implementations of these systems cannot support the continuous display of multimedia objects and suffer from frequent disruptions and delays termed hiccups. This is due to the low I/O bandwidth of the current disk technology, the high bandwidth requirement of multimedia objects, and the large size of these objects that almost always requires them to be disk resident. One approach to resolve this limitation is to decluster a multimedia object across multiple disk drives in order to employ the aggregate bandwidth of several disks to support the continuous retrieval (and display) of objects. This paper describes staggered striping as a novel technique to provide effective support for multiple users accessing the different objects in the database. Detailed simulations confirm the superiority of stagge...

By providing direct data transfer between storage and client, net-work-attached storage devices have the potential to improve scal-ability for existing distributed file systems (by removing the server as a bottleneck) and bandwidth for new parallel and distributed file systems (through network striping and more efficient data paths). Together, these advantages influence a large enough fraction of the storage market to make commodity network-attached storage fea-sible. Realizing the technology’s full potential requires careful consideration across a wide range of file system, networking and security issues. This paper contrasts two network-attached storage architectures-(l) Networked SCSI disks (NetSCSI) are network-attached storage devices with minimal changes from the familiar SCSI interface, while (2) Network-Attached Secure Disks (NASD) are drives that support independent client access to drive object services. To estimate the potential performance benefits of these architectures, we develop an analytic model and perform trace-driven replay experiments based on AFS and NFS traces. Our results suggest that NetSCSI can reduce tile server load during a burst of NFS or AFS activity by about 30%. With the NASD archi-tecture, server load (during burst activity) can be reduced by a fac-tor of up to five for AFS and up to ten for NFS. 1

Improvements in disk speeds have not kept up with improvements in processor and memory speeds. One way to correct the resulting speed mismatch is to stripe data across many disks. In this paper, we address how to stripe data to get maximum performance from the disks. Specifically, we examine how to choose the striping unit, i.e. the amount of logically contiguous data on each disk. We synthesize rules for determining the best striping unit for a given range of workloads. We show how the choice of striping unit depends on only two parameters: 1) the number of outstanding requests in the disk system at any given time, and 2) the average positioning time data transfer rate of the disks. We derive an equation for the optimal striping unit as a function of these two parameters; we also show how to choose the striping unit without prior knowledge about the workload.

The increasing performance and decreasing cost of processors and memory are causing system intelligence to move into peripherals from the CPU. Storage system designers are using this trend toward "excess" compute power to perform more complex processing and optimizations inside storage devices. To date, such optimizations have been at relatively low levels of the storage protocol. At the same time, trends in storage density, mechanics, and electronics are eliminating the bottleneck in moving data off the media and putting pressure on interconnects and host processors to move data more efficiently. We propose a system called Active Disks that takes advantage of processing power on individual disk drives to run application-level code. Moving portions of an application's processing to execute directly at disk drives can dramatically reduce data traffic and take advantage of the storage parallelism already present in large systems today. We discuss several types of appl...

We describe and evaluate a strategy for declustering the parity encoding in a redundant disk array. This declustered parity organization balances cost against data reliability and performance during failure recovery. It is targeted at highly-available parity-based arrays for use in continuousoperation systems. It improves on standard parity organizations by reducing the additional load on surviving disks during the reconstruction of a failed disk's contents. This yields higher user throughput during recovery, and/or shorter recovery time. We first address the generalized parity layout problem, basing our solution on balanced incomplete and complete block designs. A software implementation of declustering is then evaluated using a disk array simulator under a highly concurrent workload comprised of small user accesses. We show that declustered parity penalizes user response time while a disk is being repaired (before and during its recovery) less than comparable non-declustered (RAID5) ...

We consider the problem of external sorting in a shared-nothing multiprocessor. A critical step in the algorithms we consider is to determine the range of sort keys to be handled by each processor. We consider two techniques for determining these ranges of sort keys: exact splitting, using a parallel version of the algorithm proposed by Iyer, Ricard, and Varman; and probabilistic splitting, which uses sampling to estimate quantiles. We present analytic results showing that probabilistic splitting performs better than exact splitting. Finally, we present experimental results from an implementation of sorting via probabilistic splitting in the Gamma parallel database machine.

A crucial issue in the design of very large disk arrays is the protection of data against catastrophic disk failures. Although today single disks are highly reliable, when a disk array consists of 100 or 1000 disks, the probability that at least one disk will fail within a day or a week is high. In this paper, we address the problem of designing erasure-correcting binary linear codes that protect against the loss of data caused by disk failures in large disk arrays. We describe how such codes can be used to encode data in disk arrays, and give a simple method for data reconstruction. We discuss important reliability and performance constraints of these codes, and show how these constraints relate to properties of the parity check matrices of the codes. In so doing, we transform code design problems into combinatorial problems. Using this combinatorial framework, we present codes and prove they are optimal with respect to various reliability and performance constraints.

\We are developing a software system called PASSION: Parallel And Scalable Software for Input-Output which provides software support for high performance parallel I/O. PASSION provides support at the language, compiler, runtime as well as file system level. PASSION provides runtime procedures for parallel access to files (read/write), as well as for out-of-core computations. These routines can either be used together with a compiler to translate out-of-core data parallel programs written in a language like HPF, or used directly by application programmers. A number of optimizations such as Two-Phase Access, Data Sieving, Data Prefetching and Data Reuse have been incorporated in the PASSION Runtime Library for improved performance. PASSION also provides an initial framework for runtime support for out-of-core irregular problems. The goal of the PASSION compiler is to automatically translate out- of-core data parallel programs to node programs for distributed memory machines, with calls to the PASSION Runtime Library. At the language level, PASSION suggests extensions to HPF for out-of-core programs. At the file system level, PASSION provides support for buffering and prefetching data from disks. A portable parallel file system is also being developed as part of this project, which can be used across homogeneous or heterogeneous networks of workstations. PASSION also provides support for integrating data and task parallelism using parallel I/O techniques. We have used PASSION to implement a number of out-of-core applications such as a Laplace's equation solver, 2D FFT, Matrix Multiplication, LU Decomposition, image processing applications as well as unstructured mesh kernels in molecular dynamics and computational fluid dynamics. We are currently in the process of using PASSION in applications in CFD (3D turbulent flows), molecular structure calculations, seismic computations, and earth and space science applications such as Four-Dimensional Data Assimilation. PASSION is currently available on the Intel Paragon, Touchstone Delta and iPSC/860. Efforts are underway to port it to the IBM SP-1 and SP-2 using the Vesta Parallel File System.