Most systems in HPC make use of hierarchical designs that allow multiple levels of parallelism to be exploited by programmers. The use of multiple multi-core/multi-processor computers to form a computer cluster supports both fine-grain and large-grain parallel computation. Aggregate function communications provide an easy to use and efficient set of mechanisms for communicating and coordinating between processing elements, but the model originally targeted only fine grain parallel hardware. This work shows that a hierarchical implementation of aggregate functions is a viable alternative to MPI (the standard Message Passing Interface library) for programming clusters that provide both fine grain and large grain execution. Performance of a prototype implementation is evaluated and compared to that of MPI.

LARs (Line Associative Registers) are very wide tagged registers, used for both register-wide SWAR (SIMD Within a Register )operations and scalar operations on arbitrary fields. LARs include a large data field, type tags, source addresses, and a dirty bit, which allow them to not only replace both caches and registers in the conventional memory hierarchy, but improve on both their functions. This thesis details a LAR-based architecture, and describes the design of a compiler which can generate code for a LAR-based design. In particular, type conversion, alignment, and register allocation are discussed in detail.

CoyoteLab is an exploration in the use of Linux container technology as a means to simplify the way students in computing fields access and complete laboratory work in their educational career. This project provides two main benefits: creating a simple way for students to log in and access their coursework without anything more than their web browser, and providing course instructors a way to verify that assigned work is completed successfully. Thanks to advances in container technology and the advent of WebSockets, this becomes a middle layer between a WebSocket opened up on the client’s browser and the SSH daemon running in the user’s container on a remote server.

In a big organization there are many events organized every day. To know about the events, we typically need to check an events page, rely on flyers or on distributed pamphlets or through word of mouth. To register for an event a user now a days typically does this online which involves inputting user details. At the event, the user either signs a sheet of paper or enters credentials in a web page loaded on a tablet or other electronic device. Typically, this is a time-consuming process with many redundancies like entering user details every time the user wants to register for a new event and re-entering the details at the event. This project designs a system that eliminates these redundancies and improves event management.

Now-a-days the chatbot development has been moving from the field of Artificial-Intelligence labs to the desktops and mobile domain experts. In the fastest growing technology world, most smartphone users spend major time in the messaging apps such as Facebook messenger. A chatbot is a computer program that uses messaging channels to interact with users using natural Languages. Chatbot uses appropriate mapping techniques to transform user inputs into a relational database and fetch the data by calling an existing API and then sends an appropriate response to the user to drive its chats. Drawbacks include the need to learn and use chatbot specific languages such as AIML (Artificial Intelligence Markup Language), high botmaster interference, and the use of non-matured technology. In this project, Facebook messenger based chatbot is proposed to provide domain independent, an easy to use, smart, scalable, dynamic and conversational agent in order to get information about CSUSB. It has the unique functionalities which identify user interactions made by their natural language, and the flawless support of various application domains. This provides an ample of unique scalabilities and abilities that will be evaluated in the future phases of this project.

Communication has become a bottleneck for modern microprocessors and multi-core chips because metal wires don't scale. The problem becomes worse as the number of components increases and chips become bigger. Traditional Systems-on-Chips (SoCs) interconnect architectures are based on shared-bus communication, which can carry only one communication transaction at a time. This limits the communication bandwidth and scalability. Networks-on-Chip (NoC) were proposed as a promising solution for designing large and complex SoCs. The NoC paradigm provides better scalability and reusability for future SoCs, however, long-distance multi-hop communication through traditional metal wires suffers from both high latency and power consumption. A radical solution to address this challenge is to add long-range, low power, and high-bandwidth single-hop links between distant cores. The use of optical or on-chip RF wireless links has been explored in this context. However, all previous work has focused on regular mesh-based metal wire fabrics that were expanded with one or two additional link types only for long-distance communication. In this thesis we address the following main research questions to address the above-mentioned challenges: (1) What library of different link types would represent an optimum in the design space? (2) How would these links be used to design an application-specific NoC architecture? (3) How would applications use the resulting NoC architecture efficiently? We hypothesize that networks with a higher degree of heterogeneity, i.e., three or more link types, will improve the network throughput and consume less energy compared to traditional NoC architectures. In order to verify our hypothesis and to address the research challenges, we design and analyze optimal heterogeneous networks under different realistic traffic models by considering different cost and performance trade-offs in a comprehensive technology-agnostic simulation framework that uses metaheuristic optimization techniques. As opposed to related work, our heterogeneous links can be placed anywhere in the network, which allows to explore the entire search space. The resulting application-specific networks are then analyzed by using complex network techniques, such as community detection and small-worldness, to understand how heterogeneous link types are used to improve the NoCs performance and cost. Next, we use the application-specific networks as a target architecture for other applications. The goal is to evaluate the performance of our new NoCs for applications they have not been designed for by finding optimal resource allocations. Our results show that there is an optimal number of heterogeneous link types for each set of constraints and that networks with three or more heterogeneous link types provide significantly higher throughput along with lower energy consumption compared to both homogeneous link type and regular 2D mesh networks under three different traffic scenarios. Our evolved networks with three different technology-driven link types, namely metal wires, wireless, and optical links, provide 15% more throughput and fourteen times less energy consumption compared to homogeneous link type network. When ten different abstract link types are used in the design, 12% more throughput and 52% less energy consumption are obtained compared to networks with three different technology-driven link types. This shows that heterogeneous NoC designs based on traditional metal wires, wireless, and optical links, occupy a non-optimal spot in the entire design space. Our results further show that heterogeneous NoCs scale up significantly better in terms of performance and cost compared to mesh networks. We uncovered that network communities evolve robustly and that heterogeneous link types are efficiently establishing inter- and intra-subnet connections depending on their link type properties. We also show that mapping an application on our application-specific NoC architecture provides on average 45% more throughput at 70% less energy consumption compared to regular 2D mesh networks. The NoCs are therefore not only good for the application they were designed for, but for a broad range of other applications as well.

This thesis presents a new, practical approach to solve various NP-hard combinatorial problems of logic synthesis, logic programming, graph theory and related areas. A problem to be solved is polynomially time reduced to one of several generic combinatorial problems which can be expressed in the form of the Generalized Propositional Formula (GPF) : a Boolean product of clauses, where each clause is a sum of products of negated or non-negated literals.

Modern main memory is primarily built using dynamic random access memory (DRAM) chips. As DRAM chip scales to higher density, there are mainly three problems that impede DRAM scalability and performance improvement. First, DRAM refresh overhead grows from negligible to severe, which limits DRAM scalability and causes performance degradation. Second, although memory capacity has increased dramatically in past decade, memory bandwidth has not kept pace with CPU performance scaling, which has led to the memory wall problem. Third, DRAM dissipates considerable power and has been reported to account for as much as 40% of the total system energy and this problem exacerbates as DRAM scales up. To address these problems, 1) we propose Rank-level Piggyback Caching (RPC) to alleviate DRAM refresh overhead by servicing memory requests and refresh operations in parallel; 2) we propose a high performance and bandwidth efficient approach, called SELF, to breaking the memory bandwidth wall by exploiting die-stacked DRAM as a part of memory; 3) we propose a cost-effective and energy-efficient architecture for hybrid memory systems composed of high bandwidth memory (HBM) and phase change memory (PCM), called Dual Role HBM (DR-HBM). In DR-HBM, hot pages are tracked at a cost-effective way and migrated to the HBM to improve performance, while cold pages are stored at the PCM to save energy.

Digital Instrumentation and Control (I&C) systems in safety-related applications of next generation industrial automation systems require high levels of resilience against different fault classes. One of the more essential concepts for achieving this goal is the notion of resilient and survivable digital I&C systems. In recent years, self-healing concepts based on biological physiology have received attention for the design of robust digital systems. However, many of these approaches have not been architected from the outset with safety in mind, nor have they been targeted for the automation community where a significant need exists. This dissertation presents a new self-healing digital I&C architecture called BioSymPLe, inspired from the way nature responds, defends and heals: the stem cells in the immune system of living organisms, the life cycle of the living cell, and the pathway from Deoxyribonucleic acid (DNA) to protein. The BioSymPLe architecture is integrating biological concepts, fault tolerance techniques, and operational schematics for the international standard IEC 61131-3 to facilitate adoption in the automation industry. BioSymPLe is organized into three hierarchical levels: the local function migration layer from the top side, the critical service layer in the middle, and the global function migration layer from the bottom side. The local layer is used to monitor the correct execution of functions at the cellular level and to activate healing mechanisms at the critical service level. The critical layer is allocating a group of functional B cells which represent the building block that executes the intended functionality of critical application based on the expression for DNA genetic codes stored inside each cell. The global layer uses a concept of embryonic stem cells by differentiating these type of cells to repair the faulty T cells and supervising all repair mechanisms. Finally, two industrial applications have been mapped on the proposed architecture, which are capable of tolerating a significant number of faults (transient, permanent, and hardware common cause failures CCFs) that can stem from environmental disturbances and we believe the nexus of its concepts can positively impact the next generation of critical systems in the automation industry.

The complexity of distributed computing systems has raised two unprecedented challenges for system management. First, various customers need to be assured by meeting their required service-level agreements such as response time and throughput. Second, system power consumption must be controlled in order to avoid system failures caused by power capacity overload or system overheating due to increasingly high server density. However, most existing work, unfortunately, either relies on open-loop estimations based on off-line profiled system models, or evolves in a more ad hoc fashion, which requires exhaustive iterations of tuning and testing, or oversimplifies the problem by ignoring the coupling between different system characteristics (ie, response time and throughput, power consumption of different servers). As a result, the majority of previous work lacks rigorous guarantees on the performance and power consumption for computing systems, and may result in degraded overall system performance. In this thesis, we extensively study adaptive performance/power management and power-efficient performance management for distributed computing systems such as information dissemination systems, power grid management systems, and data centers, by proposing Multiple-Input-Multiple-Output (MIMO) control and hierarchical designs based on feedback control theory. For adaptive performance management, we design an integrated solution that controls both the average response time and CPU utilization in information dissemination systems to achieve bounded response time for high-priority information and maximized system throughput in an example information dissemination system. In addition, we design a hierarchical control solution to guarantee the deadlines of real-time tasks in power grid computing by grouping them based on their characteristics, respectively. For adaptive power management, we design MIMO optimal control solutions for power control at the cluster and server level and a hierarchical solution for large-scale data centers. Our MIMO control design can capture the coupling among different system characteristics, while our hierarchical design can coordinate controllers at different levels. For power-efficient performance management, we discuss a two-layer coordinated management solution for virtualized data centers. Experimental results in both physical testbeds and simulations demonstrate that all the solutions outperform state-of-the-art management schemes by significantly improving overall system performance.