High-performance computing (HPC) systems play an essential role in large-scale scientific computations. As the number of nodes in HPC systems continues to increase, their power consumption leads to larger energy costs. The energy costs pose a financial burden on maintaining HPC systems, which will be more challenging on future extreme-scale systems where the number of nodes and power consumption are expected to further grow. To support this growth, higher degrees of network and memory resource sharing are implemented, causing a substantial increase in performance variation and degradation. These challenges call for innovations in HPC system middleware that reduce energy cost without trading off performance. By taking the performance of an HPC system as a first-order constraint, this thesis establishes that HPC systems can participate in demand response programs while providing performance guarantees through a novel design of the middleware. Well-designed middleware also enables enhanced performance by mitigating resource contention induced by energy or cost restrictions. This thesis aims to realize these goals through two complementary approaches. First, this thesis proposes novel policies for HPC systems to enable their participation in emerging power markets, where participants reduce their energy costs by following market requirements. Our policies guarantee that the Quality-of-Service (QoS) of jobs does not drop below given constraints and systematically optimize cost reduction based on large deviation analysis in queueing theory. Through experiments on a real-world cluster whose power consumption is regulated to follow a dynamically changing power target, this thesis claims that HPC systems can participate in emerging power programs without violating the QoS constraints of jobs. Second, this thesis proposes novel resource management strategies to improve the performance of HPC systems. Better resource management can mitigate contention that causes performance degradation and poor system utilization. To resolve network contention, we design an intelligent job allocation policy for HPC systems that incorporate the state-of-the-art dragonfly network topology. Our allocation policy mitigates network contention, reduces network communication latency, and consequently improves the performance of the systems. As some latest HPC systems support the collection of high-granularity network performance metrics at runtime, we also propose a method to quantify the impact of network congestion and demonstrate that a network-data-driven job allocation policy improves HPC performance by avoiding network traffic hot spots. / 2022-01-18T00:00:00Z

abstract: In the recent times, traffic congestion and motor accidents have been a major problem for transportation in major cities. Intelligent Transportation Systems has the potential to be an effective solution in order to tackle this issue. Connected Autonomous Vehicles can cooperate at intersections, ramp merging, lane change and other conflicting scenarios in order to resolve the conflicts and avoid collisions with other vehicles. A lot of works has been proposed for specific scenarios such as intersections, ramp merging or lane change which partially solve the conflict resolution problem. Also, one of the major issues in autonomous decision making - deadlocks have not been considered in some of the works. The existing works either do not consider deadlocks or lack a safety proof. This thesis proposes a cooperative driving solution that provides a complete navigation, conflict resolution and deadlock resolution for connected autonomous vehicles. A graph-based model is used to resolve the deadlocks between vehicles and the responsibility sensitive safety (RSS) rules have been used in order to ensure safety of the autonomous vehicles during conflict detection and resolution. This algorithm provides a complete navigation solution for an autonomous vehicle from its source to destination. The algorithm ensures that accidents do not occur even in the worst-case scenario and the decision making is deadlock free. / Dissertation/Thesis / Masters Thesis Computer Engineering 2020

abstract: Mobile and Internet-of-Things (IoT) systems have been widely used in many aspects of human’s life. These systems are storing and operating on more and more sensitive data of users. Attackers may want to obtain the data to peek at users’ privacy or pollute the data to cause system malfunction. In addition, these systems are not user-friendly for some people such as children, senior citizens, and visually impaired users. Therefore, it is of cardinal significance to improve both security and usability of mobile and IoT systems. This report consists of four parts: one automatic locking system for mobile devices, one systematic study of security issues in crowdsourced indoor positioning systems, one usable indoor navigation system, and practical attacks on home alarm IoT systems. Chapter 1 overviews the challenges and existing solutions in these areas. Chapater 2 introduces a novel system ilock which can automatically and immediately lock the mobile devices to prevent data theft. Chapter 3 proposes attacks and countermeasures for crowdsourced indoor positioning systems. Chapter 4 presents a context-aware indoor navigation system which is more user-friendly for visual impaired people. Chapter 5 investigates some novel attacks on commercial home alarm systems. Chapter 6 concludes the report and discuss the future work. / Dissertation/Thesis / Doctoral Dissertation Electrical Engineering 2020

abstract: This thesis addresses the problem of recommending a viewpoint for aesthetic photography. Viewpoint recommendation is suggesting the best camera pose to capture a visually pleasing photograph of the subject of interest by using any end-user device such as drone, mobile robot or smartphone. Solving this problem enables to capture visually pleasing photographs autonomously in areal photography, wildlife photography, landscape photography or in personal photography. The viewpoint recommendation problem can be divided into two stages: (a) generating a set of dense novel views based on the basis views captured about the subject. The dense novel views are useful to better understand the scene and to know how the subject looks from different viewpoints and (b) each novel is scored based on how aesthetically good it is. The viewpoint with the greatest aesthetic score is recommended for capturing a visually pleasing photograph. / Dissertation/Thesis / Masters Thesis Computer Engineering 2019

abstract: Visual navigation is a multi-disciplinary field across computer vision, machine learning and robotics. It is of great significance in both research and industrial applications. An intelligent agent with visual navigation ability will be capable of performing the following tasks: actively explore in environments, distinguish and localize a requested target and approach the target following acquired strategies. Despite a variety of advances in mobile robotics, enabling an autonomous with above-mentioned abilities is still a challenging and complex task. However, the solution to the task is very likely to accelerate the landing of assistive robots. Reinforcement learning is a method that trains autonomous robot based on rewarding desired behaviors to help it obtain an action policy that maximizes rewards while the robot interacting with the environment. Through trial and error, an agent learns sophisticated and skillful strategies to handle complex tasks in the environment. Inspired by navigation procedures of human beings that when navigating through environments, humans reason about accessible spaces and geometry of the environment a lot based on first-person view, figure out the destination and then ease over, this work develops a model that maps from pixels to actions and inherently estimate the target as well as the free-space map. The model has three major constituents: (i) a cognitive mapper that maps the topologic free-space map from first-person view images, (ii) a target recognition network that locates a desired object and (iii) an action policy deep reinforcement learning network. Further, a planner model with cascade architecture based on multi-scale semantic top-down occupancy map input is proposed. / Dissertation/Thesis / Masters Thesis Computer Engineering 2019

High Performance Computing (HPC) is becoming increasingly important throughout science and engineering as ever more complex problems must be solved through computational simulations. In these large computational applications, the latency of communication between processing nodes is often the key factor that limits performance. An emerging alternative computer architecture that addresses the latency problem is the FPGA-centric cluster (FCC); in these systems, the devices (FPGAs) are directly interconnected and thus many layers of hardware and software are avoided. The result can be scalability not currently achievable with other technologies. In FCCs, FPGAs serve multiple functions: accelerator, network interface card (NIC), and router. Moreover, because FPGAs are configurable, there is substantial opportunity to tailor the router hardware to the application; previous work has demonstrated that such application-aware configuration can effect a substantial improvement in hardware efficiency. One constraint of FCCs is that it is convenient for their interconnect to be static, direct, and have a two or three dimensional mesh topology. Thus, applications that are naturally of a different dimensionality (have a different logical topology) from that of the FCC must be remapped to obtain optimal performance. In this thesis we study various aspects of the mapping problem for FCCs. There are two major research thrusts. The first is finding the optimal mapping of logical to physical topology. This problem has received substantial attention by both the theory community, where topology mapping is referred to as graph embedding, and by the High Performance Computing (HPC) community, where it is a question of process placement. We explore the implications of the different mapping strategies on communication behavior in FCCs, especially on resulting load imbalance. The second major research thrust is built around the hypothesis that applications that need to be remapped (due to differing logical and physical topologies) will have different optimal router configurations from those applications that do not. For example, due to remapping, some virtual or physical communication links may have little occupancy; therefore fewer resources should be allocated to them. Critical here is the creation of a new set of parameterized hardware features that can be configured to best handle load imbalances caused by remapping. These two thrusts form a codesign loop: certain mapping algorithms may be differentially optimal due to application-aware router reconfiguration that accounts for this mapping. This thesis has four parts. The first part introduces the background and previous work related to communication in general and, in particular, how it is implemented in FCCs. We build on previous work on application-aware router configuration. The second part introduces topology mapping mechanisms including those derived from graph embeddings and a greedy algorithm commonly used in HPC. In the third part, topology mappings are evaluated for performance and imbalance; we note that different mapping strategies lead to different imbalances both in the overall network and in each node. The final part introduces reconfigure router design that allocates resources based on different imbalance situations caused by different mapping behaviors.

Superconducting magnet systems are the enabling technology for several research fields, e.g., experimental high-energy physics and fusion. Advanced superconducting magnet systems are strongly needed to achieve ever-higher beam energy in particle accelerators. They are also extensively used in plasma confinement for fusion. The energy stored in a magnet converts to heat when the magnet is quenching, i.e., a state change from superconducting to normal. The temperature increase and the high turn-to-turn voltage developed in a quench may degrade or damage the magnet. Thus, one of the key issues for the successful operation of superconducting magnets is the quench detection and protection. This thesis discusses the self-field quench behavior of YBa₂Cu₃O₇₋δ (YBCO) coated conductors, one of the promising high-Tc conductors for superconducting magnets. The YBCO samples are provided by American Superconductor Corporation (AMSC) and SuperPower Incorporated (SPI). Samples are cryocooled and tested in self-field. A heat pulse generated by a heater fixed atop the sample is used to initiate a normal zone. Consecutive voltage taps are soldered along the sample to monitor the voltage development during a quench. Temperature profile is measured by type E thermocouples fixed along the sample. Minimum quench energy (MQE) and normal zone propagation velocity (NZPV) are measured as a function of operation temperature and transport current. It is found that the minimum quench energy (MQE) is on the order of 1 J and increases as the operation temperature decreases. MQE also increases with decreasing transport current. The normal zone propagation velocity (NZPV) is on the order of 10 mm/s and increases when the operation temperature decreases. It also increases with increasing transport current. Thus, an intrinsic trade-off exists between higher MQE (better stability) and higher NZPV (better protection performance). Lower operating temperature increases both MQE and NZPV, indicating the necessity of operating a YBCO magnet at a temperature as low as possible for better performance. Non-equipotential quench behavior is experimentally identified. When there is no obvious electrical connection between the substrate and the stabilizer in a conductor, the voltages on the substrate side rise in unison along the sample while distinct propagation and delay between the voltages traces on the stabilizer side is observed. Quench behaviors are compared between AMSC samples of similar architectures but with different stabilizers. The samples are stabilized by (1) Cu on both top and bottom sides of the sample (Cu-Cu); (2) Cu on one side and stainless steel (SS) on the other side of the sample (Cu-SS); and (3) SS on both sides of the sample (SS-SS). Quench-induced I[subscript c] degradation is observed. Quench experiments to induce degradation are conducted on AMSC's samples with three different stabilizers. A Cu-SS sample is tested at 60 K with a transport current of 30%I[subscript c]. A SS-SS sample is tested at 75 K with I[subscript t] = 26%I[subscript c]. Both samples buckled but no degradation is found. When the I[subscript t] is low, the temperature increase in the middle part of the sample is uniform and hence low spatial temperature gradient (∂T/∂x). At the same time, the temporal temperature gradient is low due to the low Joule heating rate (∂T/∂t ∼ 50 K/s) . A Cu-Cu sample is tested at 30 K with I[subscript t] = 90%I[subscript c] which is featured by a high ∂T/∂t ∼ 1800 K/s, i.e., a thermal shock. I[subscript c] of the two middle sections of the sample is degraded slightly for ∼ 3%. The threshold values for the degradation of this specific Cu-Cu sample test case are T[subscript peak] = 460 K and ∂T/∂t = 1800 K/s. ∂T/∂x is found to have less direct impact on the degradation. / A Dissertation Submitted to the Department of Eletrical and Computer Engineering in Partial FulfiLlment of the Requirements for the Degree of Doctorate of Philosophy. / Spring Semester, 2008. / October 26, 2007. / Normal Zone Propagation Velocity, YBCO Coated Conductor, Minimum Quench Energy, Stability / Includes bibliographical references. / Justin Schwartz, Professor Co-Directing Dissertation; Thomas L. Baldwin, Professor Co-Directing Dissertation; Cesar A. Luongo, Outside Committee Member; Jim P. Zheng, Committee Member.

This thesis aims to analyze neural data in an overall effort by the Charles Stark Draper Laboratory to determine an underlying pattern in brain activity in healthy individuals versus patients with a brain degenerative disorder. The neural data comes from ECoG (electrocorticography) applied to either humans or primates. Each ECoG array has electrodes that measure voltage variations which neuroscientists claim correlates to neurons transmitting signals to one another. ECoG differs from the less invasive technique of EEG (electroencephalography) in that EEG electrodes are placed above a patients scalp while ECoG involves drilling small holes in the skull to allow electrodes to be closer to the brain. Because of this ECoG boasts an exceptionally high signal-to-noise ratio and less susceptibility to artifacts than EEG [6]. While wearing the ECoG caps, the patients are asked to perform a range of different tasks. The tasks performed by patients are partitioned into different levels of mental stress i.e. how much concentration is presumably required. The specific dataset used in this thesis is derived from cognitive behavior experiments performed on primates at MGH (Massachusetts General Hospital). The content of this thesis can be thought of as a pipelined process. First the data is collected from the ECoG electrodes, then the data is pre-processed via signal processing techniques and finally the data is clustered via unsupervised learning techniques. For both the pre-processing and the clustering steps, different techniques are applied and then compared against one another. The focus of this thesis is to evaluate clustering techniques when applied to neural data. For the pre-processing step, two types of bandpass filters, a Butterworth Filter and a Chebyshev Filter were applied. For the clustering step three techniques were applied to the data, K-means Clustering, Spectral Clustering and Self-Tuning Spectral Clustering. We conclude that for pre-processing the results from both filters are very similar and thus either filter is sufficient. For clustering we conclude that K- means has the lowest amount of overlap between clusters. K-means is also the most time-efficient of the three techniques and is thus the ideal choice for this application. / 2016-10-27T00:00:00Z

Energy-related costs are among the major contributors to the total cost of ownership of data centers and high-performance computing (HPC) clusters. As a result, future data centers must be energy-efficient to meet the continuously increasing computational demand. Constraining the power consumption of the servers is a widely used approach for managing energy costs and complying with power delivery limitations. In tandem, virtualization has become a common practice, as virtualization reduces hardware and power requirements by enabling consolidation of multiple applications on to a smaller set of physical resources. However, administration and management of data center resources have become more complex due to the growing number of virtualized servers installed in data centers. Therefore, designing autonomous and adaptive energy efficiency approaches is crucial to achieve sustainable and cost-efficient operation in data centers. Many modern data centers running enterprise workloads successfully implement energy efficiency approaches today. However, the nature of multi-threaded applications, which are becoming more common in all computing domains, brings additional design and management challenges. Tackling these challenges requires a deeper understanding of the interactions between the applications and the underlying hardware nodes. Although cluster-level management techniques bring significant benefits, node-level techniques provide more visibility into application characteristics, which can then be used to further improve the overall energy efficiency of the data centers. This thesis proposes adaptive runtime power and resource management techniques on multi-core systems. It demonstrates that taking the multi-threaded workload characteristics into account during management significantly improves the energy efficiency of the server nodes, which are the basic building blocks of data centers. The key distinguishing features of this work are as follows: We implement the proposed runtime techniques on state-of-the-art commodity multi-core servers and show that their energy efficiency can be significantly improved by (1) taking multi-threaded application specific characteristics into account while making resource allocation decisions, (2) accurately tracking dynamically changing power constraints by using low-overhead application-aware runtime techniques, and (3) coordinating dynamic adaptive decisions at various layers of the computing stack, specifically at system and application levels. Our results show that efficient resource distribution under power constraints yields energy savings of up to 24% compared to existing approaches, along with the ability to meet power constraints 98% of the time for a diverse set of multi-threaded applications.

Mobile devices have become integral parts of our society. They handle our diverse computing needs from simple daily tasks (i.e., text messaging, e-mail) to complex graphics and media processing under a limited battery budget. Mobile system-on-chip (SoC) designs have become increasingly sophisticated to handle performance needs of diverse workloads and to improve user experience. Unfortunately, power and thermal constraints have also emerged as major concerns. Increased power densities and temperatures substantially impair user experience due to frequent throttling as well as diminishing device reliability and battery life. Addressing these concerns becomes increasingly challenging due to increased complexities at both hardware (e.g., heterogeneous CPUs, accelerators) and software (e.g., vast number of applications, multi-threading). Enabling sustained user experience in face of these challenges requires (1) practical runtime management solutions that can reason about the performance needs of users and applications while optimizing power and temperature; (2) tools for analyzing real-world mobile application behavior and performance. This thesis aims at improving sustained user experience under thermal limitations by incorporating insights from real-world mobile applications into runtime management. This thesis first proposes thermally-efficient and Quality-of-Service (QoS) aware runtime management techniques to enable sustained performance. Our work leverages inherent QoS tolerance of users in real-world applications and introduces QoS-temperature tradeoff as a viable control knob to improve user experience under thermal constraints. We present a runtime control framework, QScale, which manages CPU power and scheduling decisions to optimize temperature while strictly adhering to given QoS targets. We also design a framework, Maestro, which provides autonomous and application-aware management of QoS-temperature tradeoffs. Maestro uses our thermally-efficient QoS control framework, QScale, as its foundation. This thesis also presents tools to facilitate studies of real-world mobile applications. We design a practical record and replay system, RandR, to generate repeatable executions of mobile applications. RandR provides this capability by automatically reproducing non-deterministic input sources in mobile applications such as user inputs and network events. Finally, we focus on the non-deterministic executions in Android malware which seek to evade analysis environments. We propose the Proteus system to identify the instruction-level inputs that reveal analysis environments.