Browsing by Subject "Parallel processing"
Now showing 1 - 8 of 8
Results Per Page
Sort Options
Item Communication-based performance analysis of parallel processing machines(Texas Tech University, 2003-08) Parikh, SachinNot availableItem Dataflow parallelism for large scale data mining(2010-08) Daruru, Srivatsava; Ghosh, Joydeep; Marin, NenaThe unprecedented and exponential growth of data along with the advent of multi-core processors has triggered a massive paradigm shift from traditional single threaded programming to parallel programming. A number of parallel programming paradigms have thus been proposed and have become pervasive and inseparable from any large production environment. Also with the massive amounts of data available and with the ever increasing business need to process and analyze this data quickly at the minimum cost, there is much more demand for implementing fast data mining algorithms on cheap hardware. This thesis explores a parallel programming model called dataflow, the essence of which is computation organized by the flow of data through a graph of operators. This paradigm exhibits pipeline, horizontal and vertical parallelism and requires only the data of the active operators in memory at any given time allowing it to scale easily to very large datasets. The thesis describes the dataflow implementation of two data mining applications on huge datasets. We first develop an efficient dataflow implementation of a Collaborative Filtering (CF) algorithm based on weighted co-clustering and test its effectiveness on a large and sparse Netflix data. This implementation of the recommender system was able to rapidly train and predict over 100 million ratings within 17 minutes on a commodity multi-core machine. We then describe a dataflow implementation of a non-parametric density based clustering algorithm called Auto-HDS to automatically detect small and dense clusters on a massive astronomy dataset. This implementation was able to discover dense clusters at varying density thresholds and generate a compact cluster hierarchy on 100k points in less than 1.3 hours. We also show its ability to scale to millions of points as we increase the number of available resources. Our experimental results illustrate the ability of this model to “scale” well to massive datasets and its ability to rapidly discover useful patterns in two different applications.Item Design of algorithm transformations for VLSI array processing(Texas Tech University, 1986-12) Dorairaj, RavishankarThe rapid advances in the very large scale integrated (VLSI) technology has created a flurry of research in designing future computer architectures. Many methods have been developed for parallel processing of algorithms by directly mapping them onto parallel architectures. A procedure, based on the mathematical transformation of the index set and dependence vectors of an algorithm, is developed to find algorithm transformations for VLSI array processing. The algorithm is modeled as a program graph which is a directed graph. Techniques are suggested to regularize the data-flou in an algorithm, thereby minimizing the communication requirements of the architecture. We derive a set of sufficient conditions on the structure of data-flou of a class of algorithms, for the existence of valid transformations. The VLSI array is modeled as a directed graph, and the program graph is mapped onto this using the algorithm transformation.Item Development and application of a parallel chemical compositional reservoir simulator(2015-08) Behzadinasab, Masoud; Ezekoye, Ofodike A.; Sepehrnoori, Kamy, 1951-Simulation of large-scale and complicated reservoirs requires a large number of gridblocks, which requires a considerable amount of memory and is computationally expensive. One solution to remedy the computational problem is to take advantage of clusters of PCs and high-performance computing (HPC) widely available nowadays. We can run large-scale simulations faster and more efficiently by using parallel processing on these systems. In this research project, we develop a parallel version of an in-house chemical flooding reservoir simulator (UTCHEM), which is the most comprehensive chemical flooding simulator. Every physical feature of the original code has been incorporated in the parallel code. The simulation results of several case studies are compared to the original code for verification and performance of the parallelization. The efficiency of the parallelization is evaluated in terms of speedup using multiple numbers of processors. Consequently, we improve the parallel efficiency to carry out the simulations by minimizing the communications among the processors by modifying the coding. The speedup results in comparison to linear speedup (considering the ideal speedup) indicate excellent efficiency. However, using large number of processors causes the simulator speedup to deviate from linear and the efficiency to decrease. The reason for the degradation is that the time devoted to communication between the processors increases with number of processors. To the best of our knowledge, the parallel version of UTCHEM (UTCHEMP) is the first parallel chemical flooding reservoir simulator that can be effective in running large-scale cases. While it is not feasible to simulate large-scale chemical flooding reservoirs with millions of gridblocks in any serial simulator due to computer memory limitations, UTCHEMP makes simulation of such cases practical. Moreover, this parallel simulator can take advantage of multiple processors to run field-scale simulations with millions of gridblocks in few hours.Item On the parallelization of the linkage/fastlink package(Texas Tech University, 1999-12) Rai, AadityaThe remainder of this thesis is organized in the following manner. Chapter 2 gives the background informadon regarding parallelizadon concepts and tools. Chapter 3 explains profiling and analyzing a code for data dependencies. Chapter 4 presents a brief overview of Message-Passing-Interface (MPI). Chapter 5 discusses some concepts that need to be considered while achieving parallelization. Chapter 6 discusses the development ofthe parallel model and its implementadon. In Chapter 7, the experiments conducted and the results obtained are discussed at length. Chapter 8 discusses the future enhancements that could be done to extend this research work.Item Relational database applications' optimization and performance study(Texas Tech University, 1998-08) Thiruvaipati, PrashanthThe objective of the thesis is to develop efficient query processing techniques for large relational database applications since, when such applications have to process more than a million records, performance becomes a key issue. Some techniques rely upon massive hardware architectures and new database software to improve efficiency of large database systems. One of the objectives of the thesis, however, is to develop optimization techniques using existing hardware and software. Performance improvement may be achieved by the use of parallel application processes that can process different fragments of a database at the same time. Further performance improvement is achieved by using dynamic SQL and simulating an SQL outer join in the 'C programming language. Simulating the SQL function MAX and proper locking mechanism resulted in marginal performance improvement. Database design to support the use of parallel application processes and the other techniques is presented. Applications are built to test the techniques and the performance results are presented and discussed. Multiple test cases are run for each technique to ensure that the results are similar in time. For each technique, the scenarios of maximum performance improvement, the underlying mechanism, and possible limitations are discussed.Item Self-repair and adaptation in collective and parallel computational networks: a statistical approximation(Texas Tech University, 1990-08) Phoha, Vir ViranderHogg and Huberman have defined the global dynamics of a system made up of elementary computational cells which can be used to model processes such as speech and image recognition. In training a neural net model for adaptive behavior, such that a given set of inputs will result in specific outputs, Hogg and Huberman reported the so-called frustration effect, whereby outputs never converged to the desired class. Stability under parameter changes and general behavior of this model are open research issues. At a more fundamental level Hogg and Huberman hoped for the development of a theory of recognition of fuzzy inputs in such a way that the neural net parameters could be trained to produce specific responses to a desired set of training inputs. Towards such a theory, this work formulates an analytical model for approximating the outputs of the Hogg and Huberman model after k iterations through the M*N neural network. The analytical model is a best fit to the dynamic process in the sense of mean square error. Under regularity conditions such that the analytical model is a good fit, the well estabUshed theory of multivariate statistics can be used to understand the stability properties of the neural net.Item Two-dimensional image convolution by analog computation(Texas Tech University, 1997-05) Symes, Donald AllenVision computing has been a subject of tremendous research interest since the days of the original perceptron in the 1940's [McClelland 88]. The last decade or so has shown a great deal of progress in our understanding of, at least, the early stages of vision and of the underlying structures and processes that perform these early processes in biological systems. Despite the ever-increasing speed and power of digital computing systems, the consensus appears to be that any practical vision system will, necessarily, be a highly parallel, probably analog, computing system.