Search results

Title: TOWARDS COMPREHENSIVE COUNTERMEASURES AGAINST CYBER ATTACKS TO IMPROVE SYSTEM SURVIVABILITY
Creator: Wang, Li
Date: 2012-11-20, 2012-12
Description: Survivability refers to the capability of a system to ful ll its mission, in a timely manner, in the presence of attacks, failures, or...
Show moreSurvivability refers to the capability of a system to ful ll its mission, in a timely manner, in the presence of attacks, failures, or accidents. For many distributed systems, ensuring their survivability under directed attacks is critical. Tra c analysis, conducted by the attacker, could reveal the protocol being carried out by the components. Furthermore, having inferred the protocol, the attacker can use the pattern of the messages as a guide to the most critical components. In this thesis, we rst thwart these directed attacks by using message forwarding to reduce tra c di erences, thus diverge attackers from directed attack to random attack, which probabilistically prolongs the availability of important components in the system. Then, we investigate how to improve system availability when the system is under random attack. Although the attackers cannot di erentiate the di erences between critical and non-critical components, they can intelligently decide how to invest their resources by rationally selecting the number of components to attack. Under this case, how to maintain system reliability is another challenging issue. This thesis further discusses the attacker-defender problem and analyzes how to maximize system reliability under rational attacks. When one or more system processing elements are compromised by attackers, how to select applications and deploy their tasks to the remaining processing elements so that the system availability is maximized is also investigated in this thesis. To be more speci c, we assume the applications may have di erent values towards system availability and may or may not share the same composing tasks, and we presented two di erent approaches, i.e., Genetic Algorithm (GA) based approach and Max-Min-Min based approach to solving this problem. GA-based approach produces near optimal solutions and it can be used o -line when the performance is important and timing complexity is not the primary concern. While the Max-Min-Min based approach is computationally e cient and it is used when the timing is critical.
PH.D in Computer Science, December 2012
Show less

Title: RELIABILITY AND ENERGY ANALYSIS FOR EXTREME SCALE SYSTEMS
Creator: Yu, Li
Date: 2015, 2015-12
Description: Reliability and energy are two of the top major concerns in the development of today's supercomputers. To build a powerful machine while at...
Show moreReliability and energy are two of the top major concerns in the development of today's supercomputers. To build a powerful machine while at the same time satisfying reliability requirement and energy constraint, HPC scientists continue to seek a better understanding of system and component behaviors. Toward this end, modern systems are deployed with various monitoring and logging tools to track reliability and energy data during system operations. Since these data contain important information about system reliability and energy, they are valuable resources for understanding system behaviors. However, as system scale and complexity continue to grow, the process of collecting system data to extracting meaningful knowledge out of overwhelming reliability and energy data faces a number of key challenges. To address these challenges, my work consists of three parts, including data preprocessing, data analysis and advanced modeling.
Ph.D. in Computer Science, December 2015
Show less

Title: SPAM DETECTION IN SOCIAL NETWORKS: A CASE STUDY OF WEIBO
Creator: Guo, Chang
Date: 2012-07-11, 2012-07
Description: Online Social Network Service (OSNS) lead the fashion on internet nowadays[43]. Hundreds of millions people are using Facebook, Twitter,...
Show moreOnline Social Network Service (OSNS) lead the fashion on internet nowadays[43]. Hundreds of millions people are using Facebook, Twitter, MySpace and other similar OSNS all over the world[18]. In China, people use Sina Weibo, Tencent Weibo, Renren instead of Facebook and Twitter. With those miracle tools, people communicate with others far from them at real time. However, in the wrong hands, those virtual communicating services are vulnerable from being leveraged to spread harmful or unwelcome spam messages to large number of people instantly. Besides illegal advertisements, the even worse spam messages could mislead you to phishing websites, or malware downloading links. Your account may be compromised and be used by spammers to continue spreading the virus to your acquaintance. The fight with spammers has been over decades. Thousands of smart scholars has developed different strategies to auto filter most of the spam messages[10][12][18]. In E-mail system, the earliest platform leveraged by spammers and hackers, it is reported that 98% of the common spam e-mails could be identified[5]. But in OSNS, it is just the beginning of the fight. In this work, I present a further study on the behavior of spammers and spammer accounts in Sina Weibo, the most popular OSNS in China. From the data I collected, I learn the differences pattern between spammers and legitimate users and try to finally identify the spammers. I study a dataset of 220K user profile data and 2.1 million of their most recent posted tweets. My method could finally recognize 84.4% of the spammer account while the overall classification accuracy achieves 89.9%. Because this method does not rely on the content of messages but the structure and pattern of them, I believe this method should work well for other OSNS such as twitter as well.
M.S. in Computer Science, July 2012
Show less

Title: POWER PROFILING, ANALYSIS, LEARNING, AND MANAGEMENT FOR HIGH-PERFORMANCE COMPUTING
Creator: Wallace, Sean
Date: 2017, 2017-05
Description: As the field of supercomputing continues its relentless push towards greater speeds and higher levels of parallelism the power consumption of...
Show moreAs the field of supercomputing continues its relentless push towards greater speeds and higher levels of parallelism the power consumption of these large scale systems is steadily transitioning from a burden to a serious problem. While the machines are highly scaleable, the buildings, power supplies, etc. are not. Even the most power efficient systems today consume one to two megawatts per peata op/s. Multiplying that by 1,000 to reach the next generation of supercomputer (i.e., exascale) and the power necessary just to turn the machine on is simply impractical. Thus, power has become a primary design constraint for future supercomputing system designs. As such, it has become a matter of paramount importance to understand exactly how current generation systems utilize power and what implications this has on future systems. As the saying goes, you can't manage what you don't measure. This work addresses several large hurdles in fully understanding the power consumption of current systems and making actionable decisions based on this understanding. First, by leveraging environmental data collected from runs of real leadership class applications we analyze power consumption and temperature as it pertains to scale on a production IBM Blue Gene/Q supercomputer. Then, through development of a new power monitoring library, MonEQ, we quantitatively studied how power is consumed in major portions of the system (e.g., CPU, memory, etc.) through profiling of microbenchmarks. Expanding on this, we then studied how scale and network topology affect power consumption for several well-known benchmarks. Wanting to increase the effectiveness of our power monitoring library, we extended it to work with many of the most common classes of hardware available in today's HPC landscape. In doing so, we provided an in-depth analysis of what data is obtainable, what the process of obtaining it is like, and how data from different systems compares. Next, utilizing the knowledge gained from these experiences, we developed a new scheduling approach which utilizing power data can effectively keep a production system's power consumption under a user-specified power cap without modification to the applications running on the system. Finally, we extend this scheduling approach to be applicable to more than just one objective. In doing so, the scheduler can now optimize on multiple criteria instead of simply considering system utilization.
Ph.D. in Computer Science, May 2017
Show less

Title: TOWARD A NATURAL GENETIC/EVOLUTIONARY ALGORITHM FOR MULTIOBJECTIVE OPTIMIZATION
Creator: Ramasamy, Hariharane
Date: 2013, 2013-05
Description: Practical optimization problems often have multiple objectives, which are likely to conflict with each other, and have more than one optimal...
Show morePractical optimization problems often have multiple objectives, which are likely to conflict with each other, and have more than one optimal solution representing the best trade-offs among the competing objectives. Genetic algorithms, which optimize by repeatedly applying genetic operators to a population of possible solutions, have been used recently in multiobjective optimization, but often converge to a single solution that is not necessarily optimal due to lack of diversity in the population. Current multiobjective genetic and other evolutionary methods prevent this premature convergence by promoting new members that are dissimilar in parameter or objective space. A distance measure, which calculates similarities among the members in either objective or parameter space, is used to degrade the fitness of solutions when they are crowded in a small region. This process forces the algorithm to find new but distinct trade-off points in the objective or parameter space, but is computationally expensive. As the number of objectives or parameters increases, the methods fail to scale up and they deviate from the motivating concept of the genetic algorithm—natural evolution. We extend the standard genetic algorithm through two simple, yet powerful, changes motivated by natural evolution. In the first method, the algorithm, at each step, randomly or sequentially chooses one of the objectives for optimization; hence the method is called sequential extended genetic algorithm (SEGA). In the second method, a population is maintained for each objective, and crossover is performed selecting parents from across populations. This method is called parallel extended genetic algorithm (PEGA). We applied these methods to test problems from the literature, and to two well known problems, protein folding and multiple knapsack. We discovered our methods found better trade-off solutions than current multiobjective methods, without increasing computational complexity of genetic algorithms.
PH.D in Computer Science, May 2013
Show less

Title: DUAL-BASED APPROXIMATION ALGORITHMS FOR MULTIPLE NETWORK DESIGN PROBLEMS
Creator: Grimmer, Benjamin
Date: 2016, 2016-05
Description: We study a variety of NP-Complete network connectivity problems. Our pri- mary results come from a novel Dual-Based approach to approximating...
Show moreWe study a variety of NP-Complete network connectivity problems. Our pri- mary results come from a novel Dual-Based approach to approximating network de- sign problems with cut-based linear programming relaxations. This approach gives a 3=2-approximation to Minimum 2-Edge-Connected Spanning Subgraph that is equivalent to a previously proposed algorithm. One well-studied branch of network design models ad hoc networks where each node can either operate at high or low power. If we allow unidirectional links, we can formalize this into the problem Dual Power Assignment (DPA). Our Dual-Based approach gives a 3=2-approximation to DPA, improving the previous best known approximation of 11=7 1:57. Another standard network design problem is Minimum Strongly Con- nected Spanning Subgraph (MSCS). We propose a new problem generalizing MSCS and DPA called Star Strong Connectivity (SSC). Then we show that our Dual-Based approach achieves a 1.6-approximation ratio on SSC. As a result of our Dual-Based approximations, we prove new upper bounds on the integrality gaps of these problems. For completeness, we present a family of instances of MSCS (and thus SSC) with integrality gap approaching 4=3.
M.S. in Computer Science, May 2016
Show less

Title: SCALABLE INDEXING AND SEARCHING ON DISTRIBUTED FILE SYSTEMS
Creator: Ijagbone, Itua
Date: 2016, 2016-05
Description: Scientific applications and other High Performance applications generate large amounts of data. It’s said that unstructured data comprises...
Show moreScientific applications and other High Performance applications generate large amounts of data. It’s said that unstructured data comprises more than 90% of the world’s information [IDC2011], and it’s growing 60% annually [Grantz2008]. The large amounts of data generated from computation leads to data been dispersed over the file system. Problems begin to exist when we need to locate these files for later use. For small amount of files this might not be an issue but as the number of files begin to grow as well as the increase in size of these files, it becomes difficult locating these files on the file system using ordinary methods like GNU Grep [8], which is commonly used in High Performance Computing and Many-Task Computing environments. It is as a result of this problem that we have chosen this thesis to tackle the problem of finding files in a distributed system environment. Our work leverages the FusionFS [1] distributed file system and the Apache Lucene [10] centralized indexing engine as a fundamental building block. We designed and implemented a distributed search interface within the FusionFS file system that makes both indexing and searching the index across a distributed system simple. We have evaluated our system up to 64 nodes, compared it with Grep, Hadoop, and Cloudera, and have shown that FusionFS’s indexing capabilities have lower overheads and faster response times.
M.S. in Computer Science, May 2016
Show less

Title: Polymorphic Network-on-Chip Datapath Architecture for Reconfigurable Computing Machines
Creator: Weber, Joshua
Date: 2012-04-18, 2012-05
Description: Polymorphic processors have considerable advantages in performance over existing reconfigurable designs. Polymorphic processors combine the...
Show morePolymorphic processors have considerable advantages in performance over existing reconfigurable designs. Polymorphic processors combine the flexibility and ease of a general purpose processor with the performance optimizations made possible through reconfigurable arrays. Polymorphic processors provide all the ease of programming from a traditional general purpose processor while incorporating the significant performance gains that can be realized using reconfigurable arrays. Polymorphic processors can be categorized by the level of integration between the general purpose processor and the reconfigurable array. At coarse levels of integration, the processor and reconfigurable array execute independently and exchange data utilizing bus structures. These systems perform robustly for high level datadriven optimizations, allowing large segments of processing to be quickly performed on fast reconfigurable resources. However, the overhead of data transfer between the processor and array limits the benefit to fine grained optimizations. Other architectures attempt a tight coupling of reconfigurable arrays, placing them within the processor as reconfigurable coprocessors and functional units. This technique allows fine grained optimizations of small scale, highly repeated computations, but finds it difficult to replicate the gains made in large coarse grained optimizations. To achieve an even more tightly coupled design than any prior work, the fundamental architecture of the processor is changed. The datapath of the processor is eliminated and replaced with a network-on-chip communications framework. This framework connects a system of reconfigurable arrays. Some of these reconfigurable blocks are tasked with execution of standard, general purpose processor computations, emulating the standard pipeline stages of a SPARC processor. Additional reconfigurable blocks are available to the end-user to incorporate custom application specific optimizations. This new polymorphic NoC datapath (PolyNoC) processor is able to provide a more tightly integrated architecture with significant performance advantages. The PolyNoC processor is able to incorporate both fine and coarse grained optimizations, producing a polymorphic processor able to provide performance improvements for a wide range of target applications. This thesis presents the architectural design of the PolyNoC processor. The unique design constraints resulting from the use of the NoC as a datapath will be fully explored. The impact of these constraints will be incorporated into the design of a suitable NoC for the PolyNoC processor. A cycle-accurate simulator of the PolyNoC processor has been constructed. This simulator is used to examine the performance of the PolyNoC processor when executing unmodified, industry standard benchmark programs. To demonstrate the advantages of application specific extensions to the processor, accelerators are added for each benchmark. The performance of the Poly- NoC processor is promising.
Ph.D. Computer Science, May 2012
Show less

Title: THEORETICAL ANALYSIS OF REAL-TIME SCHEDULING ON RESOURCES WITH PERFORMANCE DEGRADATION AND PERIODIC REJUVENATION
Creator: Hua, Xiayu
Date: 2017, 2017-07
Description: In 1973, Liu and Layland [81] published their seminal paper on schedulability analysis of real-time system for both EDF and RM schedulers. In...
Show moreIn 1973, Liu and Layland [81] published their seminal paper on schedulability analysis of real-time system for both EDF and RM schedulers. In this work, they provide schedulability conditions and schedulability utilization bounds for both EDF and RM scheduling algorithms, respectively. In the following four decades, scheduling algorithms, utilization bounds and schedulability analyses for real-time tasks have been studied intensively. Amongst those studies, most of the research rely on a strong assumption that the performance of a computing resource does not change during its lifetime. Unfortunately, for many long standing real-time systems, such as data acquisition systems (DAQ) [74, 99], deep-space exploration programs [120, 119] and SCADA systems for power, water and other national infrastructures [121, 26], the performance of computational resources suffer notably performance degradations after a long and continuous execution period [61]. To overcome the performance degradation in long standing systems, countermeasures, which are also called system rejuvenation approaches in the literature [123, 61, 126], were introduced and studied in depth in the last two decades. Rejuvenation approaches recover system performance when being invoked and hence benefit most long standing applications [30, 102, 11, 12, 39]. However, for applications with real-time requirements, the system downtime caused by rejuvenation process, along with the decreasing performance during the system’s available time, makes the existing real-time scheduling theories difficult to be applied directly. To address this problem, this thesis studies the schedulability issues of a realtime task set running on long standing computing systems that suffers performance degradation and uses rejuvenation mechanism to recover. Our first study in the thesis focus on a simpler resource model, i.e. the periodic resource model, which only considers periodic rejuvenations. We introduce a method, i.e., Periodic Resource Integration, to combine multiple periodic resources into a single equivalent periodic resource and provide the schedulability analysis based on the combined periodic resource for real-time tasks. By integrating multiple periodic resources into one, existing real-time scheduling researches on single periodic resource can be directly applied on multiple periodic resources. In our second study, we extend the periodic resource mode to a new resource model, the P2-resource model, in our second work to characterize resources with both the performance degradation and the periodic rejuvenation. We formally define the P2-resource and analyze the schedulability of real-time task sets on a P2-resource. In particular, we first analyze the resource supply status of a given P2-resource and provide its supply bound and linear supply bound functions. We then developed the schedulability conditions for a task set running on a P2-resource with EDF or RM scheduling algorithms, respectively. We further derive utilization bounds of both EDF and RM scheduling algorithms, respectively, for schedulability test purposes. With the P2-resource model and the schedulability analysis on a single P2- resource, we further extend our work to multiple P2-resources. In this research, we 1) analyze the schedulability of a real-time task set on multiple P2-resources under fixedpriority scheduling algorithm, 2) introduce the GP-RM-P2 algorithm and 3) provide the utilization bound for this algorithm. Simulation results show that in most cases, the sufficient bounds we provide are tight. As the rejuvenation technology keeps advancing, many systems are now able to perform rejuvenations in different system layers. To accommodate this new advances, we study the schedulability conditions of a real-time task set on a single P2-resource with both cold or warm rejuvenations. We introduce a new resource model, the P2-resource with duel-level rejuvenation, i.e., P2D-resource, to accommodate this new feature. We first study the supply bound and the linear supply bound of a given P2D-resource. We then study the sufficient utilization bounds for both RM and EDF scheduling algorithms, respectively.
Ph.D. in Computer Science, July 2017
Show less

Title: AN AUTOMATED ENGINEERING PROCESS TO VERIFY THE CORRECT COORDINATION OF MULTILAYER RECOVERY
Creator: Kroculick, Joseph
Date: 2017, 2017-05
Description: Recovery is a critical function in backbone networks. The primary function of recovery is to provide connectivity regardless of which layer...
Show moreRecovery is a critical function in backbone networks. The primary function of recovery is to provide connectivity regardless of which layer recovery operates at. Another function of recovery is for all services traversing a failed link to be restored in a way that is consistent with a service user’s requirements. These requirements can include the consideration of factors such as (1) the cost of recovery, (2) the amount of traffic restored, and (3) the delay in restoring units of traffic. With more options available to recover traffic, providing an integrated recovery solution is necessary. An important force driving the evolution of network devices to transport services such as IP traffic is the layering of network resources. Layering enables networks to increase capacity by extending legacy SONET networks to interface with optical wavelengths. Inconsistent provisioning can prevent service continuity from being achieved during a failure. Continuity of service has been recognized as one key business goal. Furthermore, since recovery can occur at a different time than when it is provisioned, inconsistent provisioning is determined after the fact, with services left unrepaired, repaired unnecessarily at an extra cost, or not repaired fast enough. A network manager can check if recovery is consistent with a global perspective on how traffic should be restored by comparing the provisioning at each device against suitable properties of a formal representation. To address this issue an engineering method was developed to detect errors in provisioning automated recovery processes in multilayer and multiprotocol transport networks. This dependability assessment process (DAP) leverages inference techniques provided by Semantic Web technologies in order to detect network-device provisioning errors. Provisioning should be accompanied by methodologies, processes, and activities to ensure that it can be trusted to achieve a desired network state. The DAP takes into account unique constraints in the telecommunications domain including bottom-up evolution of physical layer technologies to provide connectivity, and lack of a universal model of network functionality. This method is applied to assessing the correctness of provisioning decisions for a protection switching application in a transport network in both the spatial and temporal domains.
Ph.D. in Computer Science, May 2017
Show less

Title: ENSURING SECURITY AND PRIVACY IN BIG DATA SHARING, TRADING, AND COMPUTING
Creator: Jung, Taheo
Date: 2017, 2017-05
Description: We have witnessed huge values of the big data in the last decade, and it is evidential that the data bring large added values to the business...
Show moreWe have witnessed huge values of the big data in the last decade, and it is evidential that the data bring large added values to the business in various areas. Owing to such opportunities, the data collection and archival became one of the most successful business strategies in the industry, and more and more user-generated data are now being acquired, stored, provisioned, and consumed nowadays. Increased collection made human being more closely involved in the life cycle of the big data characterized by the acquisition, storage, provisioning, and consumption, and larger security and privacy challenges emerged. People’s awareness of such threats led to various efforts by the governments, industry, and academia, and our efforts described in this dissertation also belong to them. We have investigated the security and privacy challenges emerging in various parts of the life cycle big data experience nowadays, and I present our major discoveries in this dissertation which are composed of three major parts: (1) security and privacy in storage of big data; (2) theoretic foundations of privacy-preserving data computing; (3) security in big data trading. We addressed new or existing security/privacy threats existing in different parts of the big data life cycle by either leveraging existing works in intelligent ways or by proposing our novel technologies. The contributions of our discoveries cam be summarized as the protection of user privacy and data security while supporting the original functionalities at negligible extra computation/communication/storage overhead.
Ph.D. in Computer Science, May 2017
Show less

Title: UNDERSTANDING VACCINATION ATTITUDES AND DETECTING SENTIMENT STIMULUS IN ONLINE SOCIAL MEDIA
Creator: Kadam, Mayuri
Date: 2017, 2017-05
Description: Vaccination being one of the most important decisions for public health, has become a debatable topic with the rise in anti-vaccination...
Show moreVaccination being one of the most important decisions for public health, has become a debatable topic with the rise in anti-vaccination sentiments in recent years. Knowing that vaccines have eradicated many endemic diseases, the rise in antivaccination sentiments jeopardizes the human health by altering the vaccine decisions. Rapidly changing information sources with the increased reach of online social media provide users with a huge amount of information and misinformation. Users exposed to these media perceive the provided information and hold an attitude towards it. Being an open platform of discussions and opinion expressions, online social media provides a great source for understanding people’s behavior. We use supervised learning for understanding the flow of vaccine sentiments and analyzing the user attitudes through online social media. In this thesis, we determine the events and incidences responsible for amplifying pro-vaccination and anti-vaccination sentiments. We investigate user behaviors and important topics of interest for these users. We develop a model for predicting a new user’s attitude utilizing that user’s recent Twitter activity.
M.S. in Computer Science, May 2017
Show less

Title: MATRIX: MANY-TASK COMPUTING EXECUTION FABRIC FOR EXTREME SCALES
Creator: Rajendran, Anupam
Date: 2013-05-01, 2013-05
Description: Scheduling large amount of jobs/tasks over large-scale distributed systems play a significant role to achieve high system utilization and...
Show moreScheduling large amount of jobs/tasks over large-scale distributed systems play a significant role to achieve high system utilization and throughput. Today’s state-of-the-art job management/scheduling systems have predominantly Master/Slaves architectures, which have inherent limitations, such as scalability issues at extreme scales (e.g. petascales and beyond) and single point failures. In designing the next-generation job management system that addresses both of these limitations, we argue that we must distribute the job scheduling and management; however, distributed job management introduces new challenges, such as non-trivial load balancing. This thesis proposes an adaptive work stealing technique to achieve distributed load balancing at extreme scales, those found in todays’ petascale systems towards tomorrow’s exascale systems. This thesis also presents the design, analysis and implementation of a distributed execution fabric called MATRIX (MAny-Task computing execution fabRIc at eXascales). MATRIX utilizes the adaptive work stealing algorithm for distributed load balancing and distributed hash tables for managing task metadata. MATRIX supports both high-performance computing (HPC) and many-task computing (MTC) workloads. We have validated it using synthetic workloads up to 4K-cores on a IBM BlueGene/P supercomputer. Results show that high efficiencies (e.g. 90%+) are possible with certain workloads. We study the performance of MATRIX in depth, including understanding the network traffic generated by the work stealing algorithm. Simulation results are presented up to 1M-node scales which show that work stealing is a scalable and efficient load balancing approach for many-core architectures to extreme-scale distributed systems.
M.S. in Computer Science, May 2013
Show less

Title: MAXIMIZATION OF SYSTEM UTILITY VALUE FOR TIME-SENSITIVE APPLICATIONS
Creator: Li, Shuhui
Date: 2014, 2014-12
Description: Many applications are time-sensitive in the sense that the usefulness or the quality of their end results depends on their completion time....
Show moreMany applications are time-sensitive in the sense that the usefulness or the quality of their end results depends on their completion time. Examples of this type of applications are threat detections in air defense systems [97], radar trackings [36, 85], mobile navigations by Google [79, 44], and online gaming by Nintendo [38], to name a few. Take the threat detection application as an example, clearly, the earlier a threat is detected, the higher utility the application provides, as earlier detection provides more time to eliminate the threat [97]. This demonstrates the time-sensitivity of its utility. Here, the term `utility' means the actual bene t that accrues from the delivery of services [16]. The dependence between an application's accrued utility and its completion time is often modeled by a Time Utility Function (TUF). Apparently, when a system has multiple time-sensitive applications competing for the resources, a question arises: how to schedule their execution orders such that the system can yield maximal accrued utility? This thesis is to address the question. In this thesis, two categories of scheduling problems for time-sensitive applications are investigated: single-task applications in uni-processor systems and parallel multi-task applications in multi-processor systems. For the rst category, a two-TUF application model with given execution time is introduced and two scheduling algorithms for this model are proposed. Di erent from the conventional one-TUF model which only considers the gain utility, the developed model can deal with both the gain and the penalty utilities. The model is further extended to cope with applications whose exact execution times are not known at a priori, rather only their probabilistic execution time distributions are known. For applications with variable execution times, the di culty is how to make judicious decisions about when to start, continue or abort the applications. For the second category, i.e., for parallel multi-task applications in multi-processor systems, di erent from the widely investigated sequential multi-task applications, a parallel multi-task application's execution can have both spatial and temporal in uence on other applications. We propose a metric to measure the spatial-temporal interference among parallel multi-task and time-sensitive applications with respect to accrued utility. Based on the metric, a 2-approximation algorithm is introduced for systems operate in discrete time domains and its lower bound of system total accrued utility value is proved. We also develop a heuristic scheduling algorithm to maximize system's total accrued utility value for continuous time systems. Finally, the thesis discusses how methodologies developed in the thesis can be applied to reduce system's operational cost without sacri cing applications' quality of service. We propose a model to bridge together two orthogonal scheduling criteria, i.e., the system operational cost and application response time, and solve the problem by transforming it to a system accrued utility value optimization problem. The research uses both theoretical and experimental approaches. Theorems and lemmas are developed to provide the foundations for our solutions, and at the same time, extensive experiments are conducted to empirically evaluate the performances of the developed solutions.
Ph.D. in Computer Science, December 2014
Show less

Title: DISTRIBUTED NOSQL STORAGE FOR EXTREME-SCALE SYSTEM SERVICES IN CLOUDS AND SUPERCOMPUTERS
Creator: Li, Tonglin
Date: 2015, 2015-12
Description: As supercomputers gain more parallelism at exponential rates, the storage infrastructure performance is increasing at a significantly lower...
Show moreAs supercomputers gain more parallelism at exponential rates, the storage infrastructure performance is increasing at a significantly lower rate due to relatively centralized management. This implies that the data management and data flow between the storage and compute resources is becoming the new bottleneck for large-scale applications. Similarly, cloud based distributed systems introduce other challenges stemming from the dynamic nature of cloud applications. This dissertation addresses several challenges on storage systems at extreme scales for supercomputers and clouds by designing and implementing a zero-hop distributed NoSQL storage system (ZHT), which has been tuned for the requirements of high-end computing systems. ZHT aims to be a building block for scalable distributed systems. The goals of ZHT are delivering high availability, good fault tolerance, light-weight design, persistence, dynamic joins and leaves, high throughput, and low latencies, at extreme scales (millions of nodes). We have evaluated ZHT’s performance under a variety of systems, ranging from a Linux cluster with 64-nodes, an Amazon EC2 virtual cluster up to 96-nodes, to an IBM Blue Gene/P supercomputer with 8K-nodes. This work also presents several real systems that have adopted ZHT as well as other NoSQL systems, namely ZHT/Q, FusionFS, IStore, MATRIX, Slurm++, Fabriq, FREIDAState, and WaggleDB, all of these real systems have been significantly simplified due to NoSQL storage systems, and have been shown to outperform other leading systems by orders of magnitude in some cases. Through our work, we have shown how NoSQL storage systems can help on both performance and scalability at large scales in such a variety of environments.
Ph.D. in Computer Science, December 2015
Show less

Title: NETWORK CONGESTION / RESOURCE ALLOCATION GAME
Creator: Shin, Junghwan
Date: 2013, 2013-12
Description: We first consider the K-user(player) resource allocation problem when the resources or strategies are associated with homogeneous functions....
Show moreWe first consider the K-user(player) resource allocation problem when the resources or strategies are associated with homogeneous functions. Further, we consider the K-user(player) matroid resource allocation problem satisfying the specified requirements of the users, which are maximal independent sets of a matroid. The objective is to choose strategies so as to minimize the average maximum cost incurred by a user where the cost of a strategy is the sum of the costs of the elements comprising the strategy. For k commodity networks with heterogeneous latency functions, we consider the price of anarchy (PoA) in multi-commodity selfish routing problems where the latency function of an edge has a heterogeneous dependency on the flow commodities, i.e. when the delay is dependent on the flow of individual commodities, rather than on the aggregate flow. Further we consider the price of anarchy (PoA) in multi-commodity atomic flows where the latency function of an edge has a heterogeneous dependency on the flow commodities, i.e. when the delay is dependent on the flow of individual commodities, rather than on the aggregate flow. Lastly, we show improved bounds on the price of anarchy for uniform latency functions where each edge of the network has the same delay function. We prove bounds on the price of anarchy for the above functions. Our bounds illustrate how the PoA is dependent on θ and the coefficients gij . At the end, we consider security aspects of network routing in a game-theoretic framework where an attacker is empowered with the ability for intrusion into edges of the network; on the other hand, the goal of the designer is to choose routing paths.
PH.D in Computer Science, December 2013
Show less

Title: TOWARD THE AUTOMATIC ORGANIZATION AND COMPREHENSION OF SOCIAL NETWORK COMMUNICATION
Creator: Platt, Alana
Date: 2013, 2013-05
Description: Social networking sites are radically transforming the way we communicate and relate to each other. They facilitate timely information...
Show moreSocial networking sites are radically transforming the way we communicate and relate to each other. They facilitate timely information exchange and give us unprecedented access to numerous sources of information on a myriad of topics. Although the information is available, there are a number of challenges that inhibit utilization of this information: Social Networks have a great volume of messages that the user must sift through to find relevant ones, messages are frequently repetitive, the information is not organized topically, and there is little context information. The information consumer (user) must take on many of the tasks traditionally performed by the information producer to get a “big picture” understanding of the topic. This thesis introduces a framework for an automated information gathering and organization system to facilitate the information consumer’s comprehension of a given topic. The framework addresses two primary components: the user interface for the system and identification of sub-topics. The framework was implemented as a research platform designed to bring these two components together and support future research in the domain.
PH.D in Computer Science, May 2013
Show less

Title: ACTIVE LEARNING WITH RICH FEEDBACK
Creator: Sharma, Manali
Date: 2017, 2017-07
Description: One of the goals of artificial intelligence is to build predictive models that can learn from examples and make predictions. Predictive models...
Show moreOne of the goals of artificial intelligence is to build predictive models that can learn from examples and make predictions. Predictive models are useful in many domains and applications such as predicting fraud in credit card transactions, predicting whether a patient has heart-disease, predicting whether an email is a spam, predicting crime, recognizing images, recognizing speech, and many more. Building predictive models often requires supervision from a human expert. Since there is a human in the loop, the supervision needs to be as resource-efficient as possible to save the human’s time, cost, and effort in providing supervision. One solution to make the supervision resource-efficient is active learning, in which the active learner interacts with the human to acquire supervision, usually in the form of labels, for a few selected examples to effectively learn a function that can be used to make predictions. In this thesis, I explore more intuitive and effective use of human supervision through richer interactions between the human expert and the learner, so that the human can understand the learner’s reasoning for querying examples, and provide information beyond just the labels for examples. Traditional active learning approaches select informative examples for labeling, but the human does not get to know why those examples are useful to the learner. While interacting with the learner to annotate examples, humans can provide rich feedback, such as provide their prior knowledge and understanding of the domain, explain certain characteristics of the data, suggest important attributes of the data, give rationales for why an example belongs to a certain category, and provide explanations by pointing out features that are indicative of certain labels. The challenge, however, is that traditional supervised learning algorithms can learn from labeled examples, but they are not equipped to readily absorb the rich feedback. In this thesis, we enable the learner to explain its reasons for selecting instances and devise novel methods to incorporate rich feedback from humans into the training of predictive models. Specifically, I build and evaluate four novel active learning frameworks to enrich the interactions between the human and learner. First, I introduce an active learning framework to reveal the learner’s perception of informative instances. Specifically, we enable the learner to provide its reasons for uncertainty on examples and utilize the learner’s perception of uncertainty to select better examples for training the predictive models. Second, I introduce a framework to enrich the interaction between the human and learner for document classification task. Specifically, we ask the human to annotate documents and provide rationales for their annotation by highlighting phrases that convinced them to choose a particular label for a document. Third, I introduce a framework to enrich the interaction between the human and learner for the aviation domain, where we ask subject matter experts to examine flights and provide rationales for why certain flights have safety concerns. Fourth, I introduce a framework to enrich the interaction between the human and learner for document classification task, where we ask humans to provide explanations for classification by highlighting phrases that reinforce their belief in the document’s label and striking-out phrases that weaken their belief in the document’s label. We show that enabling richer interactions between the human and learner and incorporating rich feedback into learning lead to more effective training of predictive models and better utilization of human supervision.
Ph.D. in Computer Science, July 2017
Show less

Title: SCALABLE RESOURCE MANAGEMENT IN CLOUD COMPUTING
Creator: Sadooghi, Iman
Date: 2017, 2017-05
Description: The exponential growth of data and application complexity has brought new challenges in the distributed computing field. Scientific...
Show moreThe exponential growth of data and application complexity has brought new challenges in the distributed computing field. Scientific applications are growing more diverse with various workloads, including traditional MPI high performance computing (HPC) to fine-grained loosely coupled many-task computing (MTC). Traditionally, these workloads have been shown to run well on supercomputers and highly-tuned HPC Clusters. The advent of Cloud computing has brought the attention of scientists to exploit these resources for scientific applications at a potentially lower cost. We investigate the nature of the cloud and its ability to run scientific applications efficiently. Delivering high throughput and low latency for the various types of workloads at large scales has driven us to design and implement new job scheduling and execution systems that are fully distributed and have the ability to run in public clouds. We discuss the design and implementation of a job scheduling and execution system (CloudKon). CloudKon is optimized to exploit the cloud resources efficiently through a variety of cloud services (Amazon SQS and DynamoDB) in order to get the best performance and utilization. It also supports various workloads including MTC and HPC applications concurrently. To further improve the performance and the flexibility of CloudKon, we designed and implemented a fully distributed message queue (Fabriq) that delivers an order of magnitude better performance than the Amazon Simple Queuing System (SQS). Designing Fabriq helped us expand our scheduling system to many other distributed system including non-Amazon clouds. Having Fabriq as a building block, we were able to design and implement a multipurpose task scheduling and execution framework (Albatross) that is able to efficiently run various types workloads at larger scales. Albatross provides data locality and task execution dependency. Those features enable Albatross to natively run MapReduce workloads. We evaluated CloudKon with synthetic MTC workloads, synthetic HPC workloads, and synthetic MapReduce applications on the Amazon AWS cloud with up to 1K instances. Fabriq was also evaluated with synthetic workloads on Amazon AWS cloud with up to 128 instances. Performance evaluations of Albatross show its ability to outperform Spark and Hadoop on different scenarios.
Ph.D. in Computer Science
Show less

Title: USER GENERATED DATA ANALYSIS AND UTILIZATION
Creator: Liu, Shizhu
Date: 2012-12-12, 2012-12
Description: Computer-mediated communication is becoming the most convenient and important way of sharing and exchanging information. The large volume and...
Show moreComputer-mediated communication is becoming the most convenient and important way of sharing and exchanging information. The large volume and diversity of user generated content as well as pervasive user opinions on the web make existing text processing methods ine cient and ine ective. Hence, there is a need for better ways of analyzing and utilizing user generated content. My thesis focuses on user generated data and is composed of two main parts: sentiment analysis and content analysis. I present a case study in which I use machine learning techniques to analyze real-world survey responses. Supervised techniques are exploited to classify customers' loyalty based on their comments and estimate a Net Promoter Score (NPS). NPS is a crucial indicator which has been used as a means of measuring survey results with a single estimator. I de ne three patterns to support generalized sentiment-bearing expression extraction, and design a set of heuristic rules to detect both explicit and implicit negations. By altering existing dependency with detected negations and generalized sentiment-bearing expressions I am able to construct more accurate sentiment features. Our results demonstrate that generalized dependency-based features are more e ective when compared to standard features. For content analysis, the thesis addresses the problem of user generated content summarization. I focus on two sub-problems: how to summarize the novel information from user generated content and how to present the evolutionary theme threads from temporal text collections with summaries. I design two speci c topic models for these two summarization tasks respectively. To discover similar and supplemental topics in user opinions with respect to the descriptive text provided by a publisher, I propose a semi-supervised generative model by casting the local publishers descriptive elds as a prior of a resembling topic. The most representative sentences in user opinions are classi ed based on their sentiment and used to construct a summary of x the comments. To track changes of topics in temporal text collections, I extend the probabilistic model to sentence level and use name entity to make the extracted theme thread easier to understand. Experimental results demonstrate the e ectiveness of the proposed models.
PH.D in Computer Science, December 2012
Show less

repository.iit

Search the repository

Pages

Pages

Enabled Filters

Refine Results

Type

Date

Academic status of creator

Subject

Creator

Rights