Abstract not found

. DDH extends the idea of dynamic hashing algorithms to distributed systems. DDH spreads data across multiple servers in a network using a novel autonomous location discovery algorithm that learns the bucket locations instead of using a centralized directory. We describe the design and implementation of the basic DDH algorithm using networked computers. Performance results show that the prototype of DDH hashing is roughly equivalent to conventional single-node hashing implementations when compared with CPU time or elapsed time. Finally, possible improvements are suggested to the basic DDH algorithm for increased reliability and robustness. 1 Introduction Rapidly plunging hardware costs and increasing performance of CPUs and networks mean that future file and database systems are likely to be constructed as networked clusters of nodes. Algorithms should be devised to work in these environments. This paper describes the design and implementation of a distributed hashing algorithm. Quick...

LH*lh is a new data structure for scalable high-performance hash les on the increasingly popular switched multicomputers, i.e., MIMD multiprocessor machines with distributed RAM memory and without shared memory. An LH*lh le scales up gracefully over available processors and the distributed memory, easily reaching Gbytes. Address calculus does not require any centralized component that could lead to a hot- spot. Access times to the le can be under a millisecond and the le can be used in parallel by several client processors. We showthe LH*lh design, and report on the performance analysis. This includes experiments on the Parsytec GC/PowerPlus multicomputer with up to 128 Power PCs and 32 MB of distributed RAM per node. We prove the e ciency of the method and justify various algorithmic choices that were made. LH*lh opens a new perspective for high-performance applications, especially for the database management of new types of data and in real-time environments.

Abstract not found

Modern applications increasingly require scalable, highly available and distributed storage systems. High-availability schemes typically deliver data despite up to n 1 simultaneous unavailabilities of the storage nodes (disks, processors with storage, or entire computers), where n is fixed. Such schemes are insufficient for scalable files, since the probability of more than n failures increases arbitrarily with file size. We propose a new schema termed LH*sa withstanding up to n simultaneous unavailabilities with n scaling with the file. We present LH*sa file manipulation and recovery algorithms. We discuss the access and storage performance, and variants tuning selected features. We show that LH*sa files may scale to any number of nodes, keeping the probability of data unavailability arbitrarily small. 1

Contents 1 Introduction 11 1.1 The Need for High Performance Databases . . . . . . . . 11 1.2 Conventional Databases . . . . . . . . . . . . . . . . . . 13 1.3 Distributed Databases . . . . . . . . . . . . . . . . . . . 13 1.4 Multidatabases . . . . . . . . . . . . . . . . . . . . . . . 14 1.5 Data Servers . . . . . . . . . . . . . . . . . . . . . . . . . 14 1.6 Parallel Data Servers . . . . . . . . . . . . . . . . . . . . 15 1.7 Database Machines . . . . . . . . . . . . . . . . . . . . . 17 1.8 Overview of Some Data Servers . . . . . . . . . . . . . . 18 1.9 Current Trends . . . . . . . . . . . . . . . . . . . . . . . 21 1.10 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 21 2 Properties of Structures for Servers 23 2.1 The Problem . . . . . . . . . . . . . . . . . . . . . . . . 23 2.2 Scalability . . . . . . . . . . . . . . . . . . . . . . . . . . 24 2.3 Distribution . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.4 Availabilit

We propose that the challenges in the design and implementation of an SDDS can be significantly eased by separating the design of the scalable high-availability part (HADRAM) from the design of the SDDS proper. We have designed and partially implemented a HADRAM system that allows measurement to prove the validity of the concept. All other highly available SDDS provide failure tolerance at the record level, whereas HADRAM provides it at the memory level. 1

LH* schema is among most studied Scalable Distributed Data Structures. LH* variants have been in particular developed for the high-availability files, capable of serving all the data despite unavailability of some storage sites. The scalable availability schemes, tolerating increasingly more failures when the file grows, are of particular importance. We present three high-availability LH* schemes using new concept of record grouping. We discuss the common building blocks and the specific features of each schema. We compare the design issues, properties and performance. 1

Abstract-We present a new dynamic hashing scheme for disk-based databases, called Multi-Directory Hashing (MDH). MDH uses multiple hash directories to access a file. The size of each hash directory grows dynamically with the file size. The advantages of MDH are enhanced concurrency, improved bucket utilization and smaller total directory size than single-directory hashing. The expected utilization of MDH increases monotonically and approaches 100 % as the number of hash directories increases. A variation of MDH, called Main Memory Multi-Directory Hashing (MM-MDH), is also described. MM-MDH achieves optimal search time in main memory databases. The performance of both methods is analyzed through theoretical and experimental results. Key words: Multi-Directory performance analysis Hashing, extendible hashing, parallel processing, main memory databases, 1.