This paper presents a novel implementation of parallel sparse matrix-matrix multiplication using distributed memory systems on heterogeneous hardware architecture. The proposed algorithm is expected to be linearly scalable up to several thousands of processors for matrices with dimensions over 106 (million). Our approach of parallelism is based on 1D decomposition and can work for both structured and unstructured sparse matrices. The storage mechanism is based on distributed hash lists, which reduces the latency for accessing and modifying an element of the product matrix, while reducing the overall merging time of the partial results computed by the processors. Theoretically, the time and space complexity of our algorithm is linearly proportional to the total number of non-zero elements in the product matrix C. The results of the performance evaluation show that the algorithm scales much better for sparse matrices with bigger dimensions. The speedup achieved using our algorithm is much better than other existing 1D algorithms. We have been able to achieve about 500 times speedup with only 672 processors. We also identified the impact of hardware architecture on scalability.

Internet of Things (IoT) is on the verge of experiencing a paradigm shift, the focus of which is the integration of people, services, context information, and things in the Connected Society, thus enabling Internet of Everything (IoE). Hundreds of billions of things will be connected to IoT/IoE by 2020. This massive immersion of things paves the way for sensing and analysing anything, anytime and anywhere. This everywhere computing coupled with Internet or web-enabled services have allowed access to a vast amount of distributed context information from heterogeneous sources. This enormous amount of context information will remain under-utilized if not properly managed. Therefore, this thesis proposes a new approach of logical-clustering as opposed to physical clustering aimed at enabling efficient context information management. However, applying this new approach requires many research challenges to be met. By adhering to a design science research method, this thesis addresses these challenges and proposes solutions to them. The thesis first outlines the architecture for realizing logical-clustering topology for which a two-tier hierarchical-distributed hash table (DHT) based system architecture and a Software Defined Networking (SDN)-like approach are utilized whereby the clustering identifications are managed on the top-level overlay (as context storage) and heterogeneous context information sources are controlled via the bottom level. The feasibility of the architecture has been proven with an ns-3 simulation tool. The next challenge is to enable scalable clustering identification dissemination, for which a distributed Publish/Subscribe (PubSub) model is developed. The massive number of immersed nodes further necessitates a dynamic self-organized system. The thesis concludes by proposing new algorithms with regard to autonomic management of IoT to bring about the self-organization. These algorithms enable to structure the logical-clustering topology in an organized way with minimal intervention from outside sources and further ensure that it evolves correctly. A distributed IoT context information-sharing platform, MediaSense, is employed and extended to prove the feasibility of the dynamic PubSub model and the correctness of self-organized algorithms and to utilize them as context storage. Promising results have provided a high number of PubSub messages per second and fast subscription matching. Self-organization further enabled logical-clustering to evolve correctly and provided results on a par with the existing MediaSense for entity joining and high discovery rates for non-concurrent entity joining. The increase in context information requires its proper management. Being able to cluster (i.e. filter) heterogeneous context information based on context similarity can help to avoid under-utilization of resources. This thesis presents an accumulation of work which can be comprehended as a step towards realizing the vision of logical-clustering topology.

Indiana University-Purdue University Indianapolis (IUPUI) / Since the use of social media has evolved, the ability of its users to bully others has increased. One of the prevalent forms of bullying is Cyberbullying, which occurs on the social media sites such as Facebook©, WhatsApp©, and Twitter©. The past decade has witnessed a growth in cyberbullying – is a form of bullying that occurs virtually by the use of electronic devices, such as messaging, e-mail, online gaming, social media, or through images or mails sent to a mobile. This bullying is not only limited to English language and occurs in other languages. Hence, it is of the utmost importance to detect cyberbullying in multiple languages. Since current approaches to identify cyberbullying are mostly focused on English language texts, this thesis proposes a new approach (called Multilingual Cyberbullying Detection System) for the detection of cyberbullying in multiple languages (English, Hindi, and Marathi). It uses two techniques, namely, Machine Learning-based and Lexicon-based, to classify the input data as bullying or non-bullying. The aim of this research is to not only detect cyberbullying but also provide a distributed infrastructure to detect bullying. We have developed multiple prototypes (standalone, collaborative, and cloud-based) and carried out experiments with them to detect cyberbullying on different datasets from multiple languages. The outcomes of our experiments show that the machine-learning model outperforms the lexicon-based model in all the languages. In addition, the results of our experiments show that collaboration techniques can help to improve the accuracy of a poor-performing node in the system. Finally, we show that the cloud-based configurations performed better than the local configurations.

Homomorphic Encryption, an encryption scheme only developed in the last five years, allows for arbitrary operations to be performed on encrypted data. Using this scheme, a user can encrypt data, and send it to an online service. The online service can then perform an operation on the data and generate an encrypted result. This encrypted result is then sent back to the user, who decrypts it. This decryption produces the same data as if the operation performed by the online service had been performed on the unencrypted data. This is revolutionary because it allows for users to rely on online services, even untrusted online services, to perform operations on their data, without the online service gaining any knowledge from their data. A prominent implementation of homomorphic encryption is HElib. While one is able to perform homomorphic encryption with this library, there are problems with it. It, like all other homomorphic encryption libraries, is slow relative to other encryption systems. Thus there is a need to speed it up. Because homomorphic encryption will be deployed on online services, many of them distributed systems, it is natural to modify HElib to utilize some of the tools that are available on them in an attempt to speed up run times. Thus two modified libraries were designed: GPUHElib, which utilizes a GPU, and DistributedHElib, which utilizes a distributed computing design. These designs were then tested against the original library to see if they provided any speed up.

DOGMA is a distributed computing architecture developed at Brigham Young University. It makes use of idle computers to provide additional computing resources to applications, similar to Seti@home. DOGMA's ability to scale to large numbers of computers is hindered by its strict client-server architecture. Recent research with DOGMA has shown that introducing localized peer-to-peer downloading abilities enhances DOGMA's performance while reducing the amount of network and server usage. This thesis proposes to further extend the peer-to-peer abilities of DOGMA to include peering client server communication by creating dynamic clusters of clients. The client clusters aggregate their communication with only one client communicating with the server directly. This further reduces the network traffic and server usage allowing more clients to connect to a single server and increasing the overall scalability of DOGMA systems.

Although the nearing end of Moore’s Law has been predicted numerous times in the past, it will eventually come to pass. In forethought of this, many modern computing systems have become increasingly complex, distributed, and parallel. As software is developed on and for these complex systems, a common API is necessary for gathering vital performance related metrics while remaining transparent to the user, both in terms of system impact and ease of use. Several distributed performance monitoring and testing systems have been proposed and implemented by both research and commercial institutions. However, most of these systems do not meet several fundamental criterion for a truly useful distributed performance monitoring system: 1) variable data delivery models, 2) security, 3) scalability, 4) transparency, 5) completeness, 6) validity, and 7) portability. This work presents dCAMP: Distributed Common API for Measuring Performance, a distributed performance framework built on top of Mark Gabel and Michael Haungs’ work with CAMP. This work also presents an updated and extended set of criterion for evaluating distributed performance frameworks and uses these to evaluate dCAMP and several related works.

Environmental management has increasingly become a participatory process. In recent times, emphasis has been placed on watershed-based solutions to remediate the problems of diffuse source pollution and to engage stakeholders in designing solutions. Water quality models are an integral part of this process; such models are often inaccessible to lay stakeholders. A review of the literature suggests that properly applied partnerships have several benefits that go beyond decision-making. Stakeholder education and enhancements to the eventual outcome from stakeholder insight and support are two such benefits. To aid engineers and scientists, who often do not interact directly with other stakeholders, several best practices were identified that may be applied to develop, manage, and evaluate stakeholder partnerships. Environmental Decision Support Systems (EDSSs) have been shown to be an effective way to promote stakeholder partnerships in environmental decision-making. Many current EDSSs were designed to be used by experts, thus limiting their effectiveness for stakeholder engagement. Often, these EDSSs, if designed for lay stakeholders, were not coupled with water quality models. To demonstrate that complex water quality models may be made accessible to stakeholders, without any significant changes to the modeling scheme, a web-based EDSS was developed for the Occoquan Reservoir, located in northern Virginia, U.S.A., and its tributary watershed. The developed EDSS may also be readily extended to other watersheds and their modeling programs. The current implementation of the EDSS enables users to modify land use and analyze simulated changes to water quality due to these modifications. A local-network server cluster, based on the Locally Distributed Simultaneous Model Execution (LDSME) framework, was also developed and served as a backend to the EDSS. The server cluster can support simultaneous execution of multiple water quality models or any other software on disparate computers. This system was employed to study pre-development and other land use modification scenarios in the Occoquan Watershed. The pre-development scenario offers an easy-to-understand and universally-applicable baseline for measuring waterbody and watershed restoration progress. It enabled computation of a measure called the "developed-excess," which is independent of local conditions and may be used for comparisons among various watershed sub-divisions or between watersheds. / Ph. D.

No description available.

Performing social network analysis (SNA) requires a set of powerful techniques to analyze structural information contained in interactions between social entities. Many SNA technologies and methodologies have been developed and have successfully provided significant insights for small-scale interactions. However, these techniques are not suitable for analyzing large social networks, which are very popular and important in various fields and have special structural properties that cannot be obtained from small networks or their analyses. There are a number of issues that need to be further studied in the design of current SNA techniques. A number of key issues can be embodied in three fundamental and critical challenges: long processing time, large computational resource requirements, and network dynamism. In order to address these challenges, we discuss an anytime-anywhere methodology based on a parallel/distributed computational framework to effectively and efficiently analyze large and dynamic social networks. In our methodology, large social networks are decomposed into intra-related smaller parts. A coarse-level of network analysis is built based on comprehensively analyzing each part. The partial analysis results are incrementally refined over time. Also, during the analyses process, network dynamic changes are effectively and efficiently adapted based on the obtained results. In order to evaluate and validate our methodology, we implement our methodology for a set of SNA metrics which are significant for SNA applications and cover a wide range of difficulties. Through rigorous theoretical and experimental analyses, we demonstrate that our anytime-anywhere methodology is / Ph. D.

Successful application of parallel high performance computing to practical problems requires overcoming several challenges. These range from the need to make sequential and parallel improvements in programs to the implementation of software tools which create an environment that aids sharing of high performance hardware resources and limits losses caused by hardware and software failures. In this thesis we describe our approach to meeting these challenges in the context of a Molecular Statics code. We describe sequential and parallel optimizations made to the code and also a suite of tools constructed to facilitate the execution of the Molecular Statics program on a network of parallel machines with the aim of increasing resource sharing, fault tolerance and availability. / Master of Science