Thesis (M.S.)--Oregon State University, 2007. / Printout. Includes bibliographical references (leaves 38-40). Also available on the World Wide Web.

Thesis (Ph. D.)--University of Canterbury, 2006. / Typescript (photocopy). Includes bibliographical references (p. 166-184). Also available via the World Wide Web.

No description available.

Individual differences among users of a hierarchical file system were investigated. The results of a first experiment indicated that psychometric tests of vocabulary and spatial visualization were the best predictors of task performance, accounting for 45% of the variance in the data. The spatial predictor was found to be the most influential. This was dramatically illustrated by the fact that, on the average, subjects with low spatial ability took twice as long to perform the task as those with high spatial ability. Surprisingly, experience alone did not predict task performance. A comparison of the frequency of command usage between subjects with high and low spatial abilities revealed that those with low spatial ability were getting lost in the hierarchical file structure. Based on the concept of visual momentum, two changes to the interface were proposed. The changes consisted of a partial map of the hierarchy and an analog indicator of current file position. A second experiment compared the performance of users with high and low spatial abilities on the old Verbal interface and the new Graphical interface. The Graphical interface resulted in changes in command usage that were consistent with the predictions of the visual momentum analysis. Although these changes in strategy resulted in a performance advantage for the Graphical interface, the relative performance difference between High and Low Spatial groups remained constant across interfaces. However, the new interface did result in a decrease in the within-group variability in performance. / Master of Science

Extreme scale data management in high performance computing requires consideration of the end-to-end scientific workflow process. Of particular importance for runtime performance, the write-read cycle must be addressed as a complete unit. Any optimization made to enhance writing performance must consider the subsequent impact on reading performance. Only by addressing the full write-read cycle can scientific productivity be enhanced. The ADIOS middleware developed as part of this thesis provides an API nearly as simple as the standard POSIX interface, but with the flexibilty to choose what transport mechanism(s) to employ at or during runtime. The accompanying BP file format is designed for high performance parallel output with limited coordination overheads while incorporating features to accelerate subsequent use of the output for reading operations. This pair of optimizations of the output mechanism and the output format are done such that they either do not negatively impact or greatly improve subsequent reading performance when compared to popular self-describing file formats. This end-to-end advantage of the ADIOS architecture is further enhanced through techniques to better enable asychronous data transports affording the incorporation of 'in flight' data processing operations and pseudo-transport mechanisms that can trigger workflows or other operations.

No description available.

Thesis: B.S., Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, 1977 / Includes bibliographical references. / by Alan Dale Siggia. / B.S. / B.S. Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science

Existing cache organization suffers from the inability to distinguish different types of localities, and non-selectively cache all data rather than making any attempt to take special advantage of the locality type. This causes unnecessary movement of data among the levels of the memory hierarchy and increases in miss ratio. In this dissertation I propose a split data cache architecture that will group memory accesses as scalar or array references according to their inherent locality and will subsequently map each group to a dedicated cache partition. In this system, because scalar and array references will no longer negatively affect each other, cache-interference is diminished, delivering better performance. Further improvement is achieved by the introduction of victim cache, prefetching, data flattening and reconfigurability to tune the array and scalar caches for specific application. The most significant contribution of my work is the introduction of novel cache architecture for embedded microprocessor platforms. My proposed cache architecture uses reconfigurability coupled with split data caches to reduce area and power consumed by cache memories while retaining performance gains. My results show excellent reductions in both memory size and memory access times, translating into reduced power consumption. Since there was a huge reduction in miss rates at L-1 caches, further power reduction is achieved by partially or completely shutting down L-2 data or L-2 instruction caches. The saving in cache sizes resulting from these designs can be used for other processor activities including instruction and data prefetching, branch-prediction buffers. The potential benefits of such techniques for embedded applications have been evaluated in my work. I also explore how my cache organization performs for non-numeric data structures. I propose a novel idea called "Data flattening" which is a profile based memory allocation technique to compress sparsely scattered pointer data into regular contiguous memory locations and explore the potentials of my proposed Spit cache organization for data treated with data flattening method.

Because of the high cost and impracticality of a high connectivity network, most recent research in transaction processing has focused on a distributed replicated database system. In such a system, multiple copies of a data item are created and stored at several sites in the network, so that the system is able to tolerate more crash and communication failures and attain higher data availability. However, the multiple copies also introduce a global inconsistency problem, especially in a partitioned network. In this dissertation a tree quorum algorithm is proposed to solve this problem, imposing a logical tree structure along with dynamic system reconfiguration on all the copies of each data item. The proposed algorithm can be viewed as a dynamic voting technique which, with the help of an appropriate concurrency control algorithm, exhibits the major advantages of quorum-based replica control algorithms and of the available copies algorithm, so that a single copy is read for a read operation and a quorum of copies is written for a write operation. In addition, read and write quorums are computed dynamically and independently. As a result expensive read operations, like those that require several copies of a data item to be read in most quorum schemes, are eliminated. Furthermore, the message costs of read and write operations are reduced by the use of smaller quorum sizes. Quorum sizes can be reduced to a constant in a lightly loaded system, and log n in a failure-free network, as well as [n +1/2] in a partitioned network in a heavily loaded system. On average, our algorithm requires fewer messages than the best known tree quorum algorithm, while still maintaining the same upper bound on quorum size. One-copy serializability is guaranteed with higher data availability and highest degree of fault tolerance (up to n - 1 site failures).

This thesis compares the file organization techniques that are implemented on two different types of computer systems, the large-scale and the small-scale. File organizations from representative computers in each class are examined in detail: the IBM System/370 (OS/370) and the Harris 1600 Distributed Processing System with the Extended Communications Operating System (ECOS). In order to establish the basic framework for comparison, an introduction to file organizations is presented. Additionally, the functional requirements for file organizations are described by their characteristics and user demands. Concluding remarks compare file organization techniques and discuss likely future developments of file systems.