Browsing by Author "Dr. Edward W. Davis, Committee Member"

Now showing 1 - 5 of 5

Implementing and Evaluating SCM Algorithms for Rate-Aware Prefetching
(2009-01-06) Kulkarni, Amit Vasant; Dr. Xiaosong Ma, Committee Chair; Dr. Vincent W. Freeh, Committee Member; Dr. Edward W. Davis, Committee Member
File system prefetching has been widely studied and used to hide high latency of disk I/O. However, there are very few algorithms that explicitly take the file access rate or burstiness into account to distribute resources, especially the prefetching memory. In this work we draw parallels between file system prefetching and the field of Supply Chain Management (SCM), particularly Inventory Theory. We further describe two very commonly used algorithms in SCM that directly address access rate and uncertainty. We also implement these prefetching algorithms in the Linux kernel and present the performance results of using these algorithms. Our results show that with these SCM-based algorithms, we can improve the throughput of standard Linux file transfer applications by up to 33% and the throughput of some server workloads (such as Video-on-Demand) by up to 41%.
Investigating Aesthetic Visualizations
(2007-01-03) Tateosian, Laura Gray; Dr. James C. Lester, Committee Member; Dr. Thomas L. Honeycutt, Committee Member; Dr. Edward W. Davis, Committee Member; Dr. Christopher G. Healey, Committee Chair
Visualizations enable scientists to inspect, interpret, and analyze large multi-dimensional data sets. Effective visualizations are designed to both orient and engage viewers by directing attention in response to a visual stimulus, and then encouraging a viewer's vision to linger at a given image location. Research into human visual perception provides information about how to orient viewers, using salient visual features, such as color, orientation, and flicker. Less is known about how to build engaging visualizations. Increasing the aesthetic merit of visualizations is a promising approach to increasing engagement. Intuition suggests that visualizations with a more aesthetic presentation style will be judged as more artistic, but this is an open problem. In this thesis, we explored an important question pertaining to creating aesthetic visualizations: Is it possible to affect the perceived artistic merit of a scientific visualization? To investigate this question, we developed three new painterly visualization techniques, designed to vary different visual qualities important to aesthetics: interpretational complexity (IC), indication and detail (ID), and visual complexity (VC). We conducted four experiments to investigate how these qualities affect the aesthetics. Observers were asked to rank IC, ID, and VC images, together with Master abstract and Impressionist paintings on five questions: artistic merit, pleasure, arousal, meaningfulness, and complexity. Although realistic Impressionist paintings consistently ranked as most artistic, computer visualizations were considered as artistic as and more pleasing than Master abstractionist artwork in certain situations. There was also a significant preference for aesthetic visualizations that used more sophisticated presentation styles. This provides strong evidence that our aesthetic techniques can increase the perceived artistic merit of a visualization, possibly leading to a significant improvement in the visualizations's ability to engage its viewers. We applied our experimental techniques to real meteorological and supernova data sets, to explore their capabilities in a real-world setting. Anecdotal feedback from a domain expert in astrophysics was strongly positive, further supporting the theory that enhancing the artistic merit of visualizations is a worthwhile contribution to the scientific community.
Nonphotorealistic Visualization of Multidimensional Datasets
(2003-01-26) Tateosian, Laura Gray; Dr. Chris G Healey, Committee Chair; Dr. Edward W. Davis, Committee Member; Dr. Robert St Amant, Committee Member
The huge quantities of data that are being recorded annually need to be organized and ana-lyzed. The datasets often consist of a large number of elements, each associated with multiple attributes. Our objective is to create effective, aesthetically appealing multidimensional visual-izations. By mapping element attributes to carefully chosen visual features, such visualizations support exploration, encourage prolonged inspection, and facilitate discovery of unexpected data characteristics and relationships. We present a new visualization technique that uses painted brushstrokes to represent data elements of large multidimensional datasets. Each element s attributes controls the visual features of one or more brushstrokes. To pursue aesthetic appeal, we draw inspiration from the Impressionist style of painting and apply rendering techniques from nonphotorealistic graphics. We construct our mappings to harness the strengths of the human visual system. The resulting displays are nonphotorealistic visualizations of the information in the datasets. Studies confirm that existing guidelines based on human visual perception apply to our painterly styles. Additional studies investigate the artistic appeal of our visualizations, along with the emotional and visual features that influence aesthetic judgments. The results of these studies are combined to build a tool which creates visualizations that are both effective and aesthetic. Finally, we apply our method to a real-world dataset.
Software Development using the Knowledge Insight Approach.
(2004-07-06) Menjoge, Zehlam; Dr. Thomas L. Honeycutt, Committee Chair; Dr. Mladen A. Vouk, Committee Member; Dr. Edward W. Davis, Committee Member
Software development processes currently in use often result in the production of low-quality software. The earlier the properties such as quality, security, safety and reliability are addressed during the software development life cycle, the lower is the development cost and the greater is the probability that the end product functions with no unintended consequences. Unclear and inadequate requirements, unexpected problems during implementation, unwise decisions made in the beginning stages of the projects are some of the reasons, due to which changes may be required in the later stages of the project. A good software development model is one which accommodates changes, has the ability to adapt well to them and minimizes budget and schedule overruns. The Knowledge Insight Model (KIM) is an iterative software development process. It is flexible enough to accommodate changes at any point during the software development life cycle. KIM is also a higher abstraction of many of the existing software development processes. It consists of four related models, namely, the Framer, the Maker, the Sharer and the Finder. The Framer is responsible for planning the course of action to be taken and defining the strategies, activities and tasks that are required for developing software. The Maker creates a design from the plan formulated by the Framer, which is implemented by the Finder. The Sharer defines the ways in which the activities of the Maker and the Finder can be tracked and monitored. In this way, it implements organized activity and separation of duties very effectively. The purpose of this thesis is to perform a comparative analysis of the Knowledge Insight Model with some other commonly used models, namely, the Waterfall Model, the Spiral model, the Rational Unified Process (RUP) and Extreme Programming (XP). These models are representatives of various classes of models. By studying each one of them, an understanding of the classes to which they belong, may be developed. The criteria for comparing them are based on the various aspects such as their structure and usability and the system's products, property and success models. This analysis serves to evaluate KIM. The results of this analysis show that the KIM is very flexible. It is also suitable for use as a software development model when evolutionary prototyping is required for all sizes of projects.
Sulfite Treatments in RDH and Kraft Pulping
(2003-03-13) Singh, Ajai; Dr. Hasan Jameel, Committee Chair; Dr. H.M. Chang, Committee Co-Chair; Dr. Edward W. Davis, Committee Member
This research is divided into two parts. First was sodium sulfite as a treatment agent in RDH and kraft pulping. Second was sodium sulfide as a treatment agent in modified kraft pulping. RDH cooks were done using air dried Southern Pine chips with the addition of sodium sulfite in different stages. Sodium sulfite charge was kept at 2% and 4% on OD chips. Improvement in yield and bleachability were not as expected, so to better understand the behaviour of sulfite, it was used as a pre-treatment chemical in modified kraft pulping. Sodium sulfite charge was varied upto 9% on OD chips in modified kraft cooks. Sodium sulfite was added in the pre-cooking stage, cooking and post-cooking stage. The sulfite charge was even split up between pre-treatment and post-treatment. Split addition of sodium sulfite gave good yield. Bleachability, however, didn't show the expected results. Results of this study indicated that RDH cooks with 4% sulfite charge in the Hot Black Liquor stage did not show any improvement on yield and bleachability. Split addition (pre-cook and post cook stages) of sulfite charge in the modified kraft cook increased the yield but no improvement in bleachability was seen. 0.1% AQ addition in the modified kraft cook improved the yield, but again no improvement in bleachability was achieved. Sulfide concentration is a crucial factor in kraft pulping. This study looks into the effects of sulfidity on the yield and kappa number in kraft pulping. Sulfidity was varied from 0-100% in all the stages in a 3-stage modified kraft pulping process. Pre-treatment of chips with sodium sulfide resulted in lower kappa number. Yield was higher for cooks at lower H-Factor. Higher sulfidity in the cooking stage gave higher yield at high H-Factor. The results obtained from the experimental and model data indicated that sodium sulfide treatment in the pre-treatment (pre-cook) stage is the most affective with regards to yield and kappa number. The sulfide concentration in the post-treatment (post-cook) stage had no substantial effect on kappa number and yield.