A flash memory device driver that supports a conventional UNIX file system transparently was designed. To avoid the limitations due to flash memory's restricted number of write cycles and its inability to be overwritten, this driver writes data to the flash memory system sequentially as a Log-structured File System (LFS) does and uses a cleaner to collect valid data blocks and reclaim invalid ones by erasing the corresponding flash memory regions. Measurements showed that the overhead of the cleaner has little effect on the performance of the prototype when utilization is low but that the effect becomes critical as the utilization gets higher, reducing the random write throughput from 222 Kbytes/s at 30% utilization to 40 Kbytes/s at 90 % utilization. The performance of the prototype in the Andrew Benchmark test is roughly equivalent to that of the 4.4BSD Pageable Memory based File System (MFS).

For reasons of simplicity and communication efficiency, a number of existing object-oriented database management systems are based on page server architectures; data pages are their minimum unit of transfer and client caching. Despite their efficiency, page servers are often criticized as being too restrictive when it comes to concurrency, as existing systems use pages as the minimum locking unit as well. In this paper we showhow to support object-level locking in a page server context. Several approaches are described, including an adaptive granularity approach that uses page-level locking for most pages but switches to object-level locking when finer-grained sharing is demanded. We study the performance of these approaches, comparing them to both a pure page server and a pure object server. For the range of workloads that we have examined, our results indicate that a page server is clearly preferable to an object server. Moreover, the adaptive page server is shown to provide very good ...

The performance and scalability of a client-server database system can be improved by employing client disks for caching. Client disk caching is particularly useful due to the lower cost per byte (compared to memory) and non-volatility of disk storage. Because of performance considerations, however, disk caching is not a straightforward extension of memory caching. In this paper, we examine the performance impacts of adding client disks to the storage hierarchy of a client-server DBMS and investigate the tradeoffs inherent in keeping a large volume of disk-cached data consistent. We describe and analyze four algorithms for managing disk caches. We also address two extensions to cache management algorithms that arise due to the performance characteristics of large disk caches: 1) the need for methods to reduce the work performed by the server for ensuring transaction durability, and 2) techniques for bringing a large disk-resident cache up-to-date after an extended off-line period. 1 In...

Introduction The work presented in this dissertation was motivated by two recent changes in technology: networks and large memories. The introduction of networks has led to a move away from centralized timesharing operating systems towards network operating systems. In these network operating systems each user has a personal highperformance workstation and communicates with other users across a network. Data that was once stored on a single set of disks in the timesharing systems is now distributed amongst the disks of several workstations. In fact, many of the workstations do not have any disk at all; the data for these diskless workstations is stored across the network on the disks of other workstations. The move towards network operating systems poses two problems: how to provide users with high performance and how to allow users to easily share data. Performance is a problem in network environments because each access of data may require both a network access and a disk acc

Existing distributed file systems are based on either a client -server model or a peer-to-peer model. We believe that the dynamic conditions of mobile computing and new classes of devices such as PDAs will no longer permit rigid adherence to either of these models. In this paper, we argue that a hybrid of the two will have to be used. We then propose a file system design based on such a model which exploits the ability of highly portable devices like PDAs to be carried by their owners at all times. We examine the relationship between mobile computing personae and file systems and show how the persona concept can be used to support the file system needs of mobile users. Finally we discuss some issues raised by the implementation we are building. 1. The Need for a Hybrid Model In the client-server model, used by distributed file systems such as AFS [3] and Coda [4], a select subset of the machines are given the special status of servers and act as the primary storage sites for files. At...

In this paper we present the behavior of PAFS in a scientific environment where big parallel applications are run. PAFS is a parallel/distributed file system with a cooperative cache that avoids the coherence problem while it continues to be highly scalable and achieves very good performance. The results presented in this paper have been obtained through simulation using the CHARISMA trace files and have been compared with other cooperative algorithms. Keywords Input/Output, Parallel/Distributed File System, Cooperative Cache, Cache Coherence, PAFS, xFS, CHARISMA. 1 Introduction In recent years, a lot of work has been devoted to parallel I/O and parallel/distributed file systems. This work has produced many different file systems along with as many different caching policies. A common exponent in most of these file systems is the idea of cooperation between nodes in order to achieve a better system performance. A good example is the idea of cooperative caches. In a cooperative cache ...

With the increasing use of mobile workstations for a wide variety of tasks and associated information needs, and with many variations of available networks, access to data becomes a prime consideration. This paper discusses issues of workstation mobility and proposes a solution wherein the data structures are accessed in an encapsulated form-through the Portable File System (PFS) wrapper. The paper discusses an implementation of the Portable File System, highlighting the architecture and commenting upon performance of an experimental system. Although investigations have been focused upon mobile access of WWW documents, this technique could be applied to any mobile data access situation.

In this paper we present PAFS, a new parallel/distributed file system. Within the whole file system, special interest is placed on the caching and prefetching mechanisms. We present a cooperative cache that avoids the coherence problem while it continues to be highly scalable and achieves very good performance. We also present an aggressive prefetching algorithm that allows full utilization of the big caches offered by the cooperative cache mechanism. Keywords Input/Output, Parallel/Distributed File System, Cooperative Cache, Aggressive Prefetching, PAFS, xFS. 1 Introduction In recent years, lots of work has been devoted to parallel I/O and parallel/distributed file systems. This work has produced many different file systems along with as many different caching policies. A common exponent in most of these file systems is the idea of cooperation between nodes in order to achieve a better system performance. A good example is the idea of cooperative caches. In a cooperative cache all ...

The proliferation of computers has made it possible to do computer-related tasks in many places. At each place a user works, he or she runs applications, specifies preferences for them, and is allowed to access resources and files according to local rules. These elements, together with the nameto -resource mappings, can be referred to as a user's computing persona. Currently, an implicit persona is created wherever a user is given access to a machine and the persona evolves as applications, preferences and resources change. This proposal outlines a method for making personae explicit and user-centric, instead of machine-centric, thus creating mobile computing personae. It also describes the underlying components that are needed to support this shift. First, it discusses a mobile file system that provides file access as the user moves as a pre-requisite technology. Second, an application-based checkpoint and restart protocol is proposed; this would enable applications to be more reliab...

The need to have a mobile user's computing persona or environment move along with him or her must be satisfied if mobile computing is to become attractive to the general user. One way to do this without depending on a universal network infrastructure is to have the user carry his or her persona on a persona carrier, a device that stays with its owner at all times. Based on this idea, we have proposed a design that allows a user to access the distributed file system he or she normally uses even at isolated computers. We have built a prototype implementation that uses Coda as the distributed file system, and we find that it performs as well as a connected Coda client. 1. Introduction There have been two broad developments in recent years that have fueled the promise of mobile computing. One has been the rapid advances in computer hardware that have resulted in computing devices becoming more powerful while simultaneously becoming more portable and affordable. The second has consisted of ...