Non-uniform data distribution for communication-efficient parallel clustering

Global communicationrequirements andloadimbalanceof someparalleldataminingalgorithms arethe major obstacles to exploitthe computational power of large-scale systems. This work investigates how non-uniform data distributions can be exploited to remove the global communication requirement and to reduce the communication costin parallel data mining algorithms and, in particular, in the k-means algorithm for cluster analysis. In the straightforward parallel formulation of the k-means algorithm, data and computation loads are uniformly distributed over the processing nodes. This approach has excellent load balancing characteristics that may suggest it could scale up to large and extreme-scale parallel computing systems. However, at each iteration step the algorithm requires a global reduction operationwhichhinders thescalabilityoftheapproach.Thisworkstudiesadifferentparallelformulation of the algorithm where the requirement of global communication is removed, while maintaining the same deterministic nature ofthe centralised algorithm. The proposed approach exploits a non-uniform data distribution which can be either found in real-world distributed applications or can be induced by means ofmulti-dimensional binary searchtrees. The approachcanalso be extended to accommodate an approximation error which allows a further reduction ofthe communication costs. The effectiveness of the exact and approximate methods has been tested in a parallel computing system with 64 processors and in simulations with 1024 processing element

Dynamic group communication for large-scale parallel data mining

Exascale systems are the next frontier in high-performance computing and are expected to deliver a performance of the order of 10^18 operations per second using massive multicore processors. Very large- and extreme-scale parallel systems pose critical algorithmic challenges, especially related to concurrency, locality and the need to avoid global communication patterns. This work investigates a novel protocol for dynamic group communication that can be used to remove the global communication requirement and to reduce the communication cost in parallel formulations of iterative data mining algorithms. The protocol is used to provide a communication-efficient parallel formulation of the k-means algorithm for cluster analysis. The approach is based on a collective communication operation for dynamic groups of processes and exploits non-uniform data distributions. Non-uniform data distributions can be either found in real-world distributed applications or induced by means of multidimensional binary search trees. The analysis of the proposed dynamic group communication protocol has shown that it does not introduce significant communication overhead. The parallel clustering algorithm has also been extended to accommodate an approximation error, which allows a further reduction of the communication costs. The effectiveness of the exact and approximate methods has been tested in a parallel computing system with 64 processors and in simulations with 1024 processing elements.

Quality of Service in Distributed Stream Processing for large scale Smart Pervasive Environments

The wide diffusion of cheap, small, and portable sensors integrated in an unprecedented large variety of devices and the availability of almost ubiquitous Internet connectivity make it possible to collect an unprecedented amount of real time information about the environment we live in. These data streams, if properly and timely analyzed, can be exploited to build new intelligent and pervasive services that have the potential of improving people's quality of life in a variety of cross concerning domains such as entertainment, health-care, or energy management. The large heterogeneity of application domains, however, calls for a middleware-level infrastructure that can effectively support their different quality requirements. In this thesis we study the challenges related to the provisioning of differentiated quality-of-service (QoS) during the processing of data streams produced in pervasive environments. We analyze the trade-offs between guaranteed quality, cost, and scalability in streams distribution and processing by surveying existing state-of-the-art solutions and identifying and exploring their weaknesses. We propose an original model for QoS-centric distributed stream processing in data centers and we present Quasit, its prototype implementation offering a scalable and extensible platform that can be used by researchers to implement and validate novel QoS-enforcement mechanisms. To support our study, we also explore an original class of weaker quality guarantees that can reduce costs when application semantics do not require strict quality enforcement. We validate the effectiveness of this idea in a practical use-case scenario that investigates partial fault-tolerance policies in stream processing by performing a large experimental study on the prototype of our novel LAAR dynamic replication technique. Our modeling, prototyping, and experimental work demonstrates that, by providing data distribution and processing middleware with application-level knowledge of the different quality requirements associated to different pervasive data flows, it is possible to improve system scalability while reducing costs.

Parallel vector processing of multidimensional orthogonal transforms for digital signal processing applications

The performance of the parallel vector implementation of the one- and two-dimensional orthogonal transforms is evaluated. The orthogonal transforms are computed using actual or modified fast Fourier transform (FFT) kernels. The factors considered in comparing the speed-up of these vectorized digital signal processing algorithms are discussed and it is shown that the traditional way of comparing th execution speed of digital signal processing algorithms by the ratios of the number of multiplications and additions is no longer effective for vector implementation; the structure of the algorithm must also be considered as a factor when comparing the execution speed of vectorized digital signal processing algorithms. Simulation results on the Cray X/MP with the following orthogonal transforms are presented: discrete Fourier transform (DFT), discrete cosine transform (DCT), discrete sine transform (DST), discrete Hartley transform (DHT), discrete Walsh transform (DWHT), and discrete Hadamard transform (DHDT). A comparison between the DHT and the fast Hartley transform is also included.(34 refs)

Design and Development of an FPGA-based Distributed Computing Processing Platform

This thesis presents two frameworks- a software framework and a hardware core manager framework- which, together, can be used to develop a processing platform using a distributed system of field-programmable gate array (FPGA) boards. The software framework providesusers with the ability to easily develop applications that exploit the processing power of FPGAs while the hardware core manager framework gives users the ability to configure and interact with multiple FPGA boards and/or hardware cores. This thesis describes the design and development of these frameworks and analyzes the performance of a system that was constructed using the frameworks. The performance analysis included measuring the effect of incorporating additional hardware components into the system and comparing the system to a software-only implementation. This work draws conclusions based on the provided results of the performance analysis and offers suggestions for future work.

Divided we stand: Parallel distributed stack memory management

We present an overview of the stack-based memory management techniques that we used in our non-deterministic and-parallel Prolog systems: &-Prolog and DASWAM. We believe that the problems associated with non-deterministic and-parallel systems are more general than those encountered in or-parallel and deterministic and-parallel systems, which can be seen as subsets of this more general case. We develop on the previously proposed "marker scheme", lifting some of the restrictions associated with the selection of goals while keeping (virtual) memory consumption down. We also review some of the other problems associated with the stack-based management scheme, such as handling of forward and backward execution, cut, and roll-backs.

Issues in implementing ACE: A stack copying based and-or parallel system

We discuss several issues involved in the implementation of ACE, a model capable of exploiting both And-parallelism and Or-parallelism in Prolog in a unified framework. The Orparallel model that ACE employs is based on the idea of stack-copying developed for Muse, while the model of independent And-parallelism is based on the distributed stack approach of &-Prolog. We discuss the organization of the workers, a number of sharing assumtions, techniques for work load detection, and issues relaed to which parts need to be copied when a flexible and-scheduling strategy is used.

An abstract machine based execution model for computer architecture design and efficient implementation of logic programs in parallel.

The term "Logic Programming" refers to a variety of computer languages and execution models which are based on the traditional concept of Symbolic Logic. The expressive power of these languages offers promise to be of great assistance in facing the programming challenges of present and future symbolic processing applications in Artificial Intelligence, Knowledge-based systems, and many other areas of computing. The sequential execution speed of logic programs has been greatly improved since the advent of the first interpreters. However, higher inference speeds are still required in order to meet the demands of applications such as those contemplated for next generation computer systems. The execution of logic programs in parallel is currently considered a promising strategy for attaining such inference speeds. Logic Programming in turn appears as a suitable programming paradigm for parallel architectures because of the many opportunities for parallel execution present in the implementation of logic programs. This dissertation presents an efficient parallel execution model for logic programs. The model is described from the source language level down to an "Abstract Machine" level suitable for direct implementation on existing parallel systems or for the design of special purpose parallel architectures. Few assumptions are made at the source language level and therefore the techniques developed and the general Abstract Machine design are applicable to a variety of logic (and also functional) languages. These techniques offer efficient solutions to several areas of parallel Logic Programming implementation previously considered problematic or a source of considerable overhead, such as the detection and handling of variable binding conflicts in AND-Parallelism, the specification of control and management of the execution tree, the treatment of distributed backtracking, and goal scheduling and memory management issues, etc. A parallel Abstract Machine design is offered, specifying data areas, operation, and a suitable instruction set. This design is based on extending to a parallel environment the techniques introduced by the Warren Abstract Machine, which have already made very fast and space efficient sequential systems a reality. Therefore, the model herein presented is capable of retaining sequential execution speed similar to that of high performance sequential systems, while extracting additional gains in speed by efficiently implementing parallel execution. These claims are supported by simulations of the Abstract Machine on sample programs.

Parallel Computational Intelligence-Based Multi-Camera Surveillance System

In this work, we present a multi-camera surveillance system based on the use of self-organizing neural networks to represent events on video. The system processes several tasks in parallel using GPUs (graphic processor units). It addresses multiple vision tasks at various levels, such as segmentation, representation or characterization, analysis and monitoring of the movement. These features allow the construction of a robust representation of the environment and interpret the behavior of mobile agents in the scene. It is also necessary to integrate the vision module into a global system that operates in a complex environment by receiving images from multiple acquisition devices at video frequency. Offering relevant information to higher level systems, monitoring and making decisions in real time, it must accomplish a set of requirements, such as: time constraints, high availability, robustness, high processing speed and re-configurability. We have built a system able to represent and analyze the motion in video acquired by a multi-camera network and to process multi-source data in parallel on a multi-GPU architecture.

The application of parallel processing methods to vibrational analysis

The trend in modal extraction algorithms is to use all the available frequency response functions data to obtain a global estimate of the natural frequencies, damping ratio and mode shapes. Improvements in transducer and signal processing technology allow the simultaneous measurement of many hundreds of channels of response data. The quantity of data available and the complexity of the extraction algorithms make considerable demands on the available computer power and require a powerful computer or dedicated workstation to perform satisfactorily. An alternative to waiting for faster sequential processors is to implement the algorithm in parallel, for example on a network of Transputers. Parallel architectures are a cost effective means of increasing computational power, and a larger number of response channels would simply require more processors. This thesis considers how two typical modal extraction algorithms, the Rational Fraction Polynomial method and the Ibrahim Time Domain method, may be implemented on a network of transputers. The Rational Fraction Polynomial Method is a well known and robust frequency domain 'curve fitting' algorithm. The Ibrahim Time Domain method is an efficient algorithm that 'curve fits' in the time domain. This thesis reviews the algorithms, considers the problems involved in a parallel implementation, and shows how they were implemented on a real Transputer network.

Analytical approximations to predict performance measures of manufacturing systems with job failures and parallel processing

Parallel processing is prevalent in many manufacturing and service systems. Many manufactured products are built and assembled from several components fabricated in parallel lines. An example of this manufacturing system configuration is observed at a manufacturing facility equipped to assemble and test web servers. Characteristics of a typical web server assembly line are: multiple products, job circulation, and paralleling processing. The primary objective of this research was to develop analytical approximations to predict performance measures of manufacturing systems with job failures and parallel processing. The analytical formulations extend previous queueing models used in assembly manufacturing systems in that they can handle serial and different configurations of paralleling processing with multiple product classes, and job circulation due to random part failures. In addition, appropriate correction terms via regression analysis were added to the approximations in order to minimize the gap in the error between the analytical approximation and the simulation models. Markovian and general type manufacturing systems, with multiple product classes, job circulation due to failures, and fork and join systems to model parallel processing were studied. In the Markovian and general case, the approximations without correction terms performed quite well for one and two product problem instances. However, it was observed that the flow time error increased as the number of products and net traffic intensity increased. Therefore, correction terms for single and fork-join stations were developed via regression analysis to deal with more than two products. The numerical comparisons showed that the approximations perform remarkably well when the corrections factors were used in the approximations. In general, the average flow time error was reduced from 38.19% to 5.59% in the Markovian case, and from 26.39% to 7.23% in the general case. All the equations stated in the analytical formulations were implemented as a set of Matlab scripts. By using this set, operations managers of web server assembly lines, manufacturing or other service systems with similar characteristics can estimate different system performance measures, and make judicious decisions - especially setting delivery due dates, capacity planning, and bottleneck mitigation, among others.

Analytical Approximations to Predict Performance Measures of Manufacturing Systems with Job Failures and Parallel Processing

Parallel processing is prevalent in many manufacturing and service systems. Many manufactured products are built and assembled from several components fabricated in parallel lines. An example of this manufacturing system configuration is observed at a manufacturing facility equipped to assemble and test web servers. Characteristics of a typical web server assembly line are: multiple products, job circulation, and paralleling processing. The primary objective of this research was to develop analytical approximations to predict performance measures of manufacturing systems with job failures and parallel processing. The analytical formulations extend previous queueing models used in assembly manufacturing systems in that they can handle serial and different configurations of paralleling processing with multiple product classes, and job circulation due to random part failures. In addition, appropriate correction terms via regression analysis were added to the approximations in order to minimize the gap in the error between the analytical approximation and the simulation models. Markovian and general type manufacturing systems, with multiple product classes, job circulation due to failures, and fork and join systems to model parallel processing were studied. In the Markovian and general case, the approximations without correction terms performed quite well for one and two product problem instances. However, it was observed that the flow time error increased as the number of products and net traffic intensity increased. Therefore, correction terms for single and fork-join stations were developed via regression analysis to deal with more than two products. The numerical comparisons showed that the approximations perform remarkably well when the corrections factors were used in the approximations. In general, the average flow time error was reduced from 38.19% to 5.59% in the Markovian case, and from 26.39% to 7.23% in the general case. All the equations stated in the analytical formulations were implemented as a set of Matlab scripts. By using this set, operations managers of web server assembly lines, manufacturing or other service systems with similar characteristics can estimate different system performance measures, and make judicious decisions - especially setting delivery due dates, capacity planning, and bottleneck mitigation, among others.