Cloud datacenters comprise hundreds or thousands of disparate application services, each having stringent performance and availability requirements, sharing a finite set of heterogeneous hardware and software resources. The implication of such complex environment is that the occurrence of performance problems, such as slow application response and unplanned downtimes, has become a norm rather than exception resulting in decreased revenue, damaged reputation, and huge human-effort in diagnosis. Though causes can be as varied as application issues (e.g. bugs), machine-level failures (e.g. faulty server), and operator errors (e.g. mis-configurations), recent studies have attributed capacity-related issues, such as resource shortage and contention, as the cause of most performance problems on the Internet today. As cloud datacenters become increasingly autonomous there is need for automated performance diagnosis systems that can adapt their operation to reflect the changing workload and topology in the infrastructure. In particular, such systems should be able to detect anomalous performance events, uncover manifestations of capacity bottlenecks, localize actual root-cause(s), and possibly suggest or actuate corrections. This thesis investigates approaches for diagnosing performance problems in cloud infrastructures. We present the outcome of an extensive survey of existing research contributions addressing performance diagnosis in diverse systems domains. We also present models and algorithms for detecting anomalies in real-time application performance and identification of anomalous datacenter resources based on operational metrics and spatial dependency across datacenter components. Empirical evaluations of our approaches shows how they can be used to improve end-user experience, service assurance and support root-cause analysis. / Cloud Control (C0590801)

The telecommunication market is growing at a very fast pace with the evolution of new technologies to support high speed throughput and the availability of a wide range of services and applications in the mobile networks. This has led to a need for communication service providers (CSPs) to shift their focus from network elements monitoring towards services monitoring and subscribers’ satisfaction by introducing the service quality management (SQM) and the customer experience management (CEM) that require fast responses to reduce the time to find and solve network problems, to ensure efficiency and proactive maintenance, to improve the quality of service (QoS) and the quality of experience (QoE) of the subscribers. While both the SQM and the CEM demand multiple information from different interfaces, managing multiple data sources adds an extra layer of complexity with the collection of data. While several studies and researches have been conducted for data analytics in mobile networks, most of them did not consider analytics based on the four dimensions involved in the mobile networks environment which are the subscriber, the handset, the service and the network element with multiple interface correlation. The main objective of this research was to develop mobile network analytics models applied to the 3G packet-switched domain by analysing data from the radio network with the Iub interface and the core network with the Gn interface to provide a fast root cause analysis (RCA) approach considering the four dimensions involved in the mobile networks. This was achieved by using the latest computer engineering advancements which are Big Data platforms and data mining techniques through machine learning algorithms. / Electrical and Mining Engineering / M. Tech. (Electrical Engineering)

Submitted by Johnny Rodrigues (johnnyrodrigues@ufcg.edu.br) on 2018-08-15T14:16:38Z No. of bitstreams: 1 ALEXANDRE NÓBREGA DUARTE - DISSERTAÇÃO PPGCC 2003..pdf: 1526817 bytes, checksum: dfc39cd8b1649bf64468cbe2eaefe99b (MD5) / Made available in DSpace on 2018-08-15T14:16:38Z (GMT). No. of bitstreams: 1 ALEXANDRE NÓBREGA DUARTE - DISSERTAÇÃO PPGCC 2003..pdf: 1526817 bytes, checksum: dfc39cd8b1649bf64468cbe2eaefe99b (MD5) Previous issue date: 2003-02-25 / Apresenta uma nova ferramenta para o diagnóstico automático de falhas em redes elétricas. A ferramenta utiliza uma técnica híbrida de correlação de eventos criada especialmente para ser utilizada em redes com constantes modificações de topologia. A técnica híbrida combina o raciocínio baseado em regras com o raciocínio baseado em modelos para eliminar as principais limitações do raciocínio baseado em regras. Com a ferramenta de diagnóstico foi possível validar o conhecimento dos especialistas em sistemas de transmissão de energia elétrica necessário para o diagnóstico de falhas em linhas de transmissão e construir uma base de regras para tal. A ferramenta foi testada no diagnóstico de falhas em linhas de transmissão de um dos cinco centros regionais da Companhia Hidro Elétrica do São Francisco (CHESF) e apresentou resultados satisfatórios de desempenho e precisão. / It presents a new tool for the automatic diagnosis of faults in electric networks. The toot uses a hybrid event correlation technique especially created to be used in networks with constant topological modifications. The hybrid technique combines ruJe-based reasoning with modelbased reasoning to eliminate the main limitations of rule-based reasoning. With the tool it was possible to validate the knowledge acquired from electric energy transmission systems specialists needed for the diagnosis of faults in transmission lines and to construct rules. The tool was tested in the diagnosis of faults in transmission lines of one of the five regional centers of the Companhia Hidro Elétrica do São Francisco (CHESF) and presented satisfactoiy results in terms of performance and precision.

Beaucoup de services vitaux de la vie quotidienne dépendent de systèmes d'ingénierie hautement complexes et interconnectés; Ces systèmes sont constitués d'un grand nombre de capteurs interconnectés, d'actionneurs et de composants du système. L'étude des systèmes interconnectés joue un rôle important dans l'étude de la fiabilité des systèmes dynamiques; car elle permet d'étudier les propriétés d'un système interconnecté en analysant ses sous-composants moins complexes. Le diagnostic des pannes est essentiel pour assurer des opérations sûres et fiables des systèmes de contrôle interconnectés. Dans toutes les situations, le système global et / ou chaque sous-système peuvent être analysés à différents niveaux pour déterminer la fiabilité du système global. Dans certains cas, il est important de déterminer les informations anormales des variables internes du sous-système local, car ce sont les causes qui contribuent au fonctionnement anormal du processus global. Cette thèse porte sur les défis de l'application de la théorie inverse du système et des techniques FDD basées sur des modèles pour traiter le problème articulaire du diagnostic des fautes et de l'analyse des causes racines (FD et RCA). Nous étudions ensuite le problème de l'inversibilité de la gauche, de l'observabilité et de la diagnosticabilité des fauts du système interconnecté, formant un algorithme FD et RCA multi-niveaux basé sur un modèle. Ce système de diagnostic permet aux composants individuels de surveiller la dynamique interne localement afin d'améliorer l'efficacité du système et de diagnostiquer des ressources de fautes potentielles pour localiser un dysfonctionnement lorsque les performances du système global se dégradent. Par conséquent, un moyen d'une combinaison d'intelligence locale avec une capacité de diagnostic plus avancée pour effectuer des fonctions FDD à différents niveaux du système est fourni. En conséquence, on peut s'attendre à une amélioration de la localisation des fauts et à de meilleurs moyens de maintenance prédictive. La nouvelle structure du système, ainsi que l'algorithme de diagnostic des fautes, met l'accent sur l'importance de la RCA de défaut des dispositifs de terrain, ainsi que sur l'influence de la dynamique interne locale sur la dynamique globale. Les contributions de cette thèse sont les suivantes: Tout d'abord, nous proposons une structure de système non linéaire interconnecté inversible qui garantit le fauts dans le sous-système de périphérique de terrain affecte la sortie mesurée du système global de manière unique et distincte. Une condition nécessaire et suffisante est développée pour assurer l'inversibilité du système interconnecté qui nécessite l'inversibilité de sous-systèmes individuels. Deuxièmement, un observateur interconnecté à deux niveaux est développé; Il se compose de deux estimateurs d'état, vise à fournir des estimations précises des états de chaque sous-système, ainsi que l'interconnexion inconnue. En outre, il fournira également une condition initiale pour le reconstructeur de données et le filtre de fauts local une fois que la procédure FD et RCA est déclenchée par tout fauts. D'une part, la mesure utilisée dans l'estimateur de l'ancien sous-système est supposée non accessible; La solution est de la remplacer par l'estimation fournie par l'estimateur de ce dernier sous-système. / Many of the vital services of everyday life depend on highly complex and interconnected engineering systems; these systems consist of large number of interconnected sensors, actuators and system components. The study of interconnected systems plays a significant role in the study of reliability theory of dynamic systems, as it allows one to investigate the properties of an interconnected system by analyzing its less complicated subcomponents. Fault diagnosis is crucial in achieving safe and reliable operations of interconnected control systems. In all situations, the global system and/or each subsystem can be analyzed at different levels in investigating the reliability of the overall system; where different levels mean from system level down to the subcomponent level. In some cases, it is important to determine the abnormal information of the internal variables of local subsystem, in order to isolate the causes that contribute to the anomalous operation of the overall process. For example, if a certain fault appears in an actuator, the origin of that malfunction can have different causes: zero deviation, leakage, clogging etc. These origins can be represented as root cause of an actuator fault. This thesis concerns with the challenges of applying system inverse theory and model based FDD techniques to handle the joint problem of fault diagnosis & root cause analysis (FD & RCA) locally and performance monitoring globally. By considering actuator as individual dynamic subsystem connected with process dynamic subsystem in cascade, we propose an interconnected nonlinear system structure. We then investigate the problem of left invertibility, fault observability and fault diagnosability of the interconnected system, forming a novel model based multilevel FD & RCA algorithm. This diagnostic algorithm enables individual component to monitor internal dynamics locally to improve plant efficiency and diagnose potential fault resources to locate malfunction when operation performance of global system degrades. Hence, a means of acombination of local intelligence with a more advanceddiagnostic capability (combining fault monitoring anddiagnosis at both local and global levels) to performFDDfunctions on different levels of the plantis provided. As a result, improved fault localization and better predictive maintenance aids can be expected. The new system structure, together with the fault diagnosis algorithm, is the first to emphasize the importance of fault RCA of field devices, as well as the influences of local internal dynamics on the global dynamics. The developed model based multi-level FD & RCA algorithm is then a first effort to combine the strength of the system level model based fault diagnosis with the component level model based fault diagnosis. The contributions of this thesis include the following: Firstly, we propose a left invertible interconnected nonlinear system structure which guarantees that fault occurred in field device subsystem will affect the measured output of the global system uniquely and distinguishably. A necessary and sufficient condition is developed to ensure invertibility of the interconnected system which requires invertibility of individual subsystems. Second, a two level interconnected observer is developed which consists of two state estimators, aims at providing accurately estimates of states of each subsystem, as well as the unknown interconnection. In addition, it will also provide initial condition for the input reconstructor and local fault filter once FD & RCA procedure is triggered by any fault. Two underlyingissues are worth to be highlighted: for one hand, the measurement used in the estimator of the former subsystem is assumed not accessible; the solution is to replace it by the estimate provided by the estimator of the latter subsystem. In fact, this unknown output is the unknown interconnection of the interconnected system, and also the input of the latter subsystem.

Following the significant transition from the traditional production industry to an informationbased economy, the telecommunications industry was faced with an explosion of innovation, resulting in a continuous change in user behaviour. The industry has made efforts to adapt to a more datadriven future, which has given rise to larger and more complex systems. Therefore, troubleshooting systems such as anomaly detection and root cause analysis are essential features for maintaining service quality and facilitating daily operations. This study aims to explore the possibilities, benefits, and drawbacks of implementing cluster analysis for anomaly detection in hybrid fibercoaxial networks. Based on the literature review on unsupervised anomaly detection and an assumption regarding the anomalous behaviour in hybrid fibercoaxial network data, the kmeans, SelfOrganizing Map, and Gaussian Mixture Model were implemented both with and without Principal Component Analysis. Analysis of the results demonstrated an increase in performance for all models when the Principal Component Analysis was applied, with kmeans outperforming both SelfOrganizing Map and Gaussian Mixture Model. On this basis, it is recommended to apply Principal Component Analysis for clusteringbased anomaly detection. Further research is necessary to identify whether cluster analysis is the most appropriate unsupervised anomaly detection approach. / Följt av övergången från den traditionella tillverkningsindustrin till en informationsbaserad ekonomi stod telekommunikationsbranschen inför en explosion av innovation. Detta skifte resulterade i en kontinuerlig förändring av användarbeteende och branschen tvingades genomgå stora ansträngningar för att lyckas anpassa sig till den mer datadrivna framtiden. Större och mer komplexa system utvecklades och således blev felsökningsfunktioner såsom anomalidetektering och rotfelsanalys centrala för att upprätthålla servicekvalitet samt underlätta för den dagliga driftverksamheten. Syftet med studien är att utforska de möjligheterna, för- samt nackdelar med att använda klusteranalys för anomalidetektering inom HFC- nätverk. Baserat på litteraturstudien för oövervakad anomalidetektering samt antaganden för anomalibeteenden inom HFC- data valdes algritmerna k- means, Self- Organizing Map och Gaussian Mixture Model att implementeras, både med och utan Principal Component Analysis. Analys av resultaten påvisade en uppenbar ökning av prestanda för samtliga modeller vid användning av PCA. Vidare överträffade k- means, både Self- Organizing Maps och Gaussian Mixture Model. Utifrån resultatanalysen rekommenderas det således att PCA bör tillämpas vid klusterings- baserad anomalidetektering. Vidare är ytterligare forskning nödvändig för att avgöra huruvida klusteranalys är den mest lämpliga metoden för oövervakad anomalidetektering.

Объект исследования – кибер-физические системы, подверженные различным инцидентам, отказам и сбоям в своей работе. Цель работы – разработка модели машинного обучения для определения корневых причин сбоев в производственной системе, а также исследование возможности использования машинного обучения для определения причин будущих сбоев. Методы исследования: сбор, анализ и синтез данных, сравнение, обобщение, классификация, аналогия, эксперимент, измерение, описание. Результаты работы: разработана и обучена модель машинного обучения для анализа корневых причин инцидентов производственной установки методом классификации на выбранном наборе «сырых» данных небольшого объема с последующей проверкой качества ее работы на тестовых данных. Область применения – обучение модели корневым причинам инцидентов (отказов, сбоев) производственных систем на имеющихся данных с последующим оперативным обнаружением причин аномальной работы систем в тандеме с работой алгоритма по автоматическому обнаружению и прогнозированию аномалий. / The object of research is cyber-physical systems that are susceptible to various incidents, failures and malfunctions in their operation. The goal of the work is to develop a machine learning model to determine the root causes of failures in a production system, as well as to explore the possibility of using machine learning to determine the causes of future failures. Research methods: collection, analysis and synthesis of data, comparison, generalization, classification, analogy, experiment, measurement, description. Results of the work: a machine learning model was developed and trained to analyze the root causes of incidents in a production facility using the classification method on a selected set of small-volume “raw” data, followed by checking the quality of its work on test data. Scope of application: training a model for the root causes of incidents (failures, failures) of production systems using available data, followed by prompt detection of the causes of abnormal operation of systems in tandem with the work of the algorithm for automatic detection and prediction of anomalies.

Service-learning, with a longstanding history in American higher education (Burkhardt & Pasque, 2005), includes three key tenets: superior academic learning, meaningful and relevant community service, and persistent civic learning (McGoldrick and Ziegert, 2002). The Carnegie Foundation for the Advancement of Teaching has created an elective classification system – Carnegie Community Engagement Classification – for institutions of higher education to demonstrate the breadth and depth of student involvement and learning through partnerships and engagement in the community (Dalton & Crosby, 2011; Hurtado & DeAngelo, 2012; Kuh et al., 2008; Pryor, Hurtado, Saenz, Santos, & Korn, 2007). Community engagement "is in the culture, commonly understood practices and knowledge, and (CCEC helps determine) whether it is really happening – rhetoric versus reality" (J. Saltmarsh, personal communication, August 11, 2014). The study considers the applications of three Carnegie Community Engagement Classification designated institutions to understand the institutionalization of service-learning over time by examining the 2008 designation and 2015 reclassification across institution types – a Private Liberal Arts College, a Private Teaching University, and a Public Research University located in the same metropolitan area. Organizational Change Theory was used as a theoretical model. Case study methodology was used in the present qualitative research to perform document analysis with qualitative interviews conducted to elucidate the data from the 2008 and 2015 CCEC applications from the three institutions. Using intra- and inter-comparative analysis, this study highlights approaches, policies, ethos, and emerging concepts to inform how higher education institutions increase the quality and quantity of service-learning opportunities that benefit higher education practitioners as well as community leaders.