Context. In recent years there is a rapid increase in demand for cloud services, as cloud computing has become a flexible platform for hosting micro services over the Internet.~Micro services are the core elements of service oriented architecture (SOA) that facilitate the deployment of distributed software systems.  Objectives. This thesis work aims at developing a typical service architecture to facilitate the deployment of compute and I/O intensive services. The thesis work also aims at evaluating the service times of these services when their respective sub services are deployed in heterogeneous environments with various loads. Methods. The thesis work has been carried out using an experimental test bed in order to evaluate the performance. The transport level performance metric called Response time is measured. It is the time taken by the server to serve the request sent by the client. Experiments have been conducted based on the objectives that are to be achieved. Results. The results obtained from the experimentation contain the average service times of a service when it is deployed on both virtual and non-virtual environments. The virtual environment is provided by Docker containers. They also include the variation in the position of their sub services. Conclusions. From results, it can be concluded that the total service times obtained are less in the case of non-virtual environment when compared to container environment.

In recent years there is a rapid increase in demand for cloud services, as cloud computing has become a flexible platform for hosting microservices over the Internet. Microservices are the core elements of service oriented architecture (SOA) that facilitate the deployment of distributed software systems. As the user requires good quality of service the response time of microservices is critical in assessing the performance of the application from the end user perspective.This thesis work aims at developing a typical service architecture to facilitate the deployment of compute and I/O intensive services. The work also aims at evaluating the service times of these service when their respective subservices are deployed in heterogeneous environments with various loads.The research work has been carried out using an experimental testbed in order to evaluate the performance. The transport level performance metric called Response time is measured. It is the time taken by the server to serve the request sent by the client. Experiments have been conducted based on the objectives that are to be achieved.The results obtained from the experimentation contain the average service times of a service when it is deployed on both virtual and non-virtual environment. The virtual environment is provided by Docker containers. They also include the variation in position of their subservices. From results it can be concluded that the total service times obtained are less in case of non-virtual environments when compared to container environment.

Thesis (M.S.)--University of Wisconsin--Madison, 1967. / eContent provider-neutral record in process. Description based on print version record. Includes bibliographical references.

Existing multi-echelon inventory optimization models and formulas were studied to get an understanding of how safety stock levels are determined. Because of the restrictive distribution assumptions of the existing safety stock formula, which are not necessarily realistic in practice, a method to analyze the performance of this formula in a more realistic setting was desired. A SimPy simulation model was designed and implemented for a simple two-stage supply chain as a way to test the performance of the safety stock formula. This implementation produced results which led to the conclusion that the safety stock formula tends to underestimate the level of safety stock needed to provide a certain service level when predicted standard deviation of demand is underestimated and the assumptions of normally distributed demand and normally distributed lead times are not fulfilled.

This bachelor degree project presents an extension of a base optimization model for a transit line which can be used to evaluate the efficiency of different configurations of a platoon with different scenarios of berths. Furthermore, different levels of autonomous vehicles are studied, three cases are presented. The first case implies that every vehicle has a driver, the second, semi-autonomous vehicles are used in the platoon which has a leading vehicle with driver. Then, the fully autonomous vehicles represent the last studied case. A new method to compute the service time in the stops which differentiate the time that passengers are boarding or alighting from delays or time lost in queues that may appear with an increasing demand is added to the base model. It is introduced also a two-step non-linear approach to the crowding factor that consider the sharp deterioration when the load factor of the bus is almost one. In this project the bus capacity has been considered as a variable to see if there is an optimum vehicle size that cover different values of demand. Numerical results are provided and the result show that vehicle platooning with equal number of vehicles than stop berths is always competitive in high-demands. Moreover, if semi-autonomous case is found the bus platooning gain effectiveness and is competitive with lower demand values. In the case of fully autonomous vehicles the gain of bus platooning is not as high as in the semiautonomous but has still an improvement and is competitive with medium demand values.

Submitted by Filipe dos Santos (fsantos@pucsp.br) on 2016-11-21T12:07:05Z No. of bitstreams: 1 Carlos Alexandre Hees.pdf: 2969591 bytes, checksum: 8695f4c272cf1dea7e363c8cd23e023c (MD5) / Made available in DSpace on 2016-11-21T12:07:05Z (GMT). No. of bitstreams: 1 Carlos Alexandre Hees.pdf: 2969591 bytes, checksum: 8695f4c272cf1dea7e363c8cd23e023c (MD5) Previous issue date: 2016-09-28 / This thesis is the result of a research and reflective analysis on social security legislation and its amendments in the political, economic and dynamic demographic distribution context. It is a descriptive qualitative research, conducted through a literature methodological approach under the historical-evolutionary perspective, through documentary research It uses the application of inductive reasoning to the standards introduced in the parental order. The aging population has led to complex discussions about the pension financing system and the prohibited social retrogression, especially after changes advent in Brazilian legislation, such as the statute 13.183 / 2015, which implements the progressive 85/95 formula, modifying the retirement calculation for service time of the General Social Security - RGPS. At this point, a close look at the effectiveness in providing this benefit before prohibited retrocession seal and demographic transition is relevant. Chapter I began studying and analyzing the evolution of social security, to its present context in Brazil. It stands the issue of social risk as a key to understand the subject. In Chapter II, it is realized the study of implications related to demography, the aging population and public policies to meet the basic needs of this segment. Therefore, in Chapter III, it passes to diagram the legal principle of conceptualization and the forbidden social retrogression as a Constitutional principle, its presence in rules, foreign doctrine, and parental jurisprudential understanding. In Chapter IV, it is analyzed progression of the 85/95 formula in social security legislation and its consequences for elderly, facing principle of forbidden social retrogression. The objective of this thesis defines itself and develops, in its own essence, the 85/95 rule analysis in Brazilian Social Security legislation, on the principle of social retrogression prohibition and population aging. The academic study is justified because the issue of social retrogression sealing is still controversial, notably by divergence of doctrine as the foundation, content and scope of the principle, reflecting the reduced number of judicial decisions that include principle of prohibited social retrogression. And facing all the study and reflection on the progress of the new retirement calculation rule, in accordance with Law 13.183 / 2015, it is concluded that the changes do not offend principle of forbidden social retrogression and presents some key elements in calculation of retirement to the harmonic development of Brazilian demography / Esta tese é resultado do trabalho de pesquisa e análise reflexiva da legislação previdenciária e suas modificações num contexto de transformações políticas, econômicas e de mudanças na distribuição demográfica. Trata-se de uma pesquisa qualitativa descritiva, conduzida por meio de uma abordagem metodológica bibliográfica, sob os prismas histórico-evolutivo e contextual-atual, por meio de pesquisa documental. Utiliza-se a aplicação do método indutivo ao arrazoar as normas introduzidas no ordenamento pátrio. O envelhecimento da população tem acarretado complexas discussões sobre o financiamento do sistema previdenciário e o debate sobre a vedação do retrocesso social, notadamente após o advento das alterações na legislação pátria, como a promulgação da Lei n.º 13.183/2015, que implanta a regra progressiva da fórmula 85/95, demudando o cálculo da aposentadoria por tempo de contribuição do Regime Geral de Previdência Social - RGPS. Portanto, ante esse quadro, é pertinente um olhar minucioso sobre a efetividade na prestação desse benefício diante da vedação de retrocesso e da transição demográfica. O Capítulo I inicia o estudo analisando a evolução da previdência social, até seu contexto atual no Brasil. Destaca-se nessa exposição a questão do risco social como fundamental para compreensão do tema. No Capítulo II, faz-se o estudo das implicações relacionadas com a demografia, o envelhecimento populacional e as políticas públicas voltadas para atender as necessidades fundamentais desse segmento. Por conseguinte, no Capítulo III, passa-se a delinear a conceituação de princípio jurídico bem como tratase a vedação do retrocesso como princípio do Estado de Direito, sua presença nas normas e doutrina estrangeiro e depois no ordenamento e no entendimento jurisprudencial pátrio. No Capítulo IV, analisa-se a progressão da fórmula 85/95 na legislação previdenciária e os reflexos dessa fórmula para o idoso à luz do princípio da vedação ao retrocesso social. O objetivo desta tese delimita-se e desenvolve-se, em sua essência, na análise da inserção da regra 85/95 na legislação da Previdência Social no Brasil, diante do princípio da proibição do retrocesso social e do envelhecimento populacional. O estudo acadêmico é justificado pela relevância do tema, porquanto a questão da vedação ao retrocesso social é ainda polêmica, notadamente pela divergência da doutrina quanto ao fundamento, conteúdo e alcance do princípio, refletindo na reduzida quantidade de decisões judiciais que abarcam o princípio da vedação ao retrocesso social. E, diante de todo o estudo e reflexão acerca da progressão da nova regra de cálculo da aposentadoria, nos termos da Lei n.º 13.183/2015, conclui-se que as alterações não ofendem o princípio da vedação ao retrocesso social e apresenta alguns elementos determinantes no cálculo de aposentadoria harmônico à evolução da demografia brasileira

This master's thesis investigates the impact of unbalanced variability on lead time using a simulation-based optimization approach. Based on the hypothesis, variation of service time has a strong effect on lead time. It has also been hypothesized that placing the high variable station, in terms of CVp, at the end of the line will increase lead time. To evaluate these hypotheses, Fact Analyzer Simulation Software version beta7 was used to simulate and optimize two different models. First, the effect of an imbalanced line in terms of service time on total lead time in a simple production line hypothetical model was investigated. In the second part of this master thesis, a real health care model was adopted from Frandsen and Engqvist’s project at Skaraborg Hospital (SkaS). By optimizing this model, using NSGA_II Algorithms, the effect of variance and mean of service time on lead time variance and mean were evaluated and compared. As a result of both the hypothetical and health care models, the total lead time will not decrease by reducing the variance of service time, indicating that the hypothesis should be rejected.    Keywords: service time, CVp, lead time, imbalanced, variability, mean lead time, variance lead time, waiting time, variability

The support of multimedia traffic over IEEE 802.11 wireless local area networks (WLANs) has recently received considerable attention. This dissertation has proposed a new framework that provides efficient channel access, service differentiation and statistical QoS guarantees in the enhanced distributed channel access (EDCA) protocol of IEEE 802.11e. In the first part of the dissertation, the new framework to provide QoS support in IEEE 802.11e is presented. The framework uses three independent components, namely, a core MAC layer, a scheduler, and an admission control. The core MAC layer concentrates on the channel access mechanism to improve the overall system efficiency. The scheduler provides service differentiation according to the weights assigned to each Access Category (AC). The admission control provides statistical QoS guarantees. The core MAC layer developed in this dissertation employs a P-Persistent based MAC protocol. A weight-based fair scheduler to obtain throughput service differentiation at each node has been used. In wireless LANs (WLANs), the MAC protocol is the main element that determines the efficiency of sharing the limited communication bandwidth of the wireless channel. In the second part of the dissertation, analytical Markov chain models for the P-Persistent 802.11 MAC protocol under unsaturated load conditions with heterogeneous loads are developed. The Markov models provide closed-form formulas for calculating the packet service time, the packet end-to-end delay, and the channel capacity in the unsaturated load conditions. The accuracy of the models has been validated by extensive NS2 simulation tests and the models are shown to give accurate results. In the final part of the dissertation, the admission control mechanism is developed and evaluated. The analytical model for P-Persistent 802.11 is used to develop a measurement-assisted model-based admission control. The proposed admission control mechanism uses delay as an admission criterion. Both distributed and centralized admission control schemes are developed and the performance results show that both schemes perform very efficiently in providing the QoS guarantees. Since the distributed admission scheme control does not have a complete state information of the WLAN, its performance is generally inferior to the centralized admission control scheme. The detailed performance results using the NS2 simulator have demonstrated the effectiveness of the proposed framework. Compared to 802.11e EDCA, the scheduler consistently achieved the desired throughput differentiation and easy tuning. The core MAC layer achieved better delays in terms of channel access, average packet service time and end-to-end delay. It also achieved higher system throughput than EDCA for any given service differentiation ratio. The admission control provided the desired statistical QoS guarantees.

No / Cognitive Radio (CR) is an emerging technology used to significantly improve the efficiency of spectrum utilization. Although some spectrum bands in the primary user's licensed spectrum are intensively used, most of the spectrum bands remain underutilized. The introduction of open spectrum and dynamic spectrum access lets the secondary (unlicensed) users, supported by cognitive radios; opportunistically utilize the unused spectrum bands. However, if a primary user returns to a band occupied by a secondary user, the occupied spectrum band is vacated immediately by handing off the secondary user's call to another idle spectrum band. Multiple spectrum handoffs can severely degrade quality of service (QoS) for the interrupted users. To avoid multiple handoffs, when a licensed primary user appears at the engaged licensed band utilized by a secondary user, an effective spectrum handoff procedure should be initiated to maintain a required level of QoS for secondary users. In other words, it enables the channel clearing while searching for target vacant channel(s) for completing unfinished transmission. This paper proposes prioritized proactive spectrum handoff decision schemes to reduce the handoff delay and the total service time. The proposed schemes have been modeled using a preemptive resume priority (PRP) M/G/1 queue to give a high priority to interrupted users to resume their transmission ahead of any other uninterrupted secondary user. The performance of proposed handoff schemes has been evaluated and compared against the existing spectrum handoff schemes. Experimental results show that the schemes developed here outperform the existing schemes in terms of average handoff delay and total service time under various traffic arrival rates as well as service rates.

Dans cette thèse, nous commençons par l'étude de la prédiction de délai d'attente des clients dans les centres d'appels multi-compétences. Le temps d'attente a un impact important sur la qualité du service perçue par les clients. L'annonce du délai d'attente permet de réduire l'incertitude du client à propos de son délai d'attente. Elle peut également augmenter la satisfaction du client et réduire le nombre d'abandons. Ceci nécessite d'avoir un bon prédicteur de délai. Malheureusement les prédicteurs existants ne sont pas adaptés pour les centres d'appels multi-compétences. Nous proposons trois types de prédicteurs qui utilisent l'apprentissage machine: le premier utilise la régression par les splines cubiques, le second emploie les réseaux de neurones artificiels, et le dernier utilise le krigeage stochastique. Les prédicteurs prennent en entrée le temps d'attente du dernier client de même type à entrer en service, la période d'arrivée du nouveau client, le nombre d'agents des groupes, la longueur de la file des clients de même type, et les longueurs des files d'attente des types servis par les mêmes agents. Ces prédicteurs donnent de bons résultats pour les systèmes multi-compétences, mais un inconvénient est qu'ils ont un grand nombre de paramètres qui doivent être appris à l'avance durant une phase d'entraînement du modèle qui nécessite une grande quantité de données et temps de calcul. Nous proposons ensuite deux nouveaux prédicteurs de délai qui sont très simples à mettre en œuvre, requièrent peu d'effort d'optimisation, ne nécessitent pas de données, et qui sont applicables dans les centres d'appels multi-compétences. Ils sont basés sur l'historique des temps d'attente des clients. Le premier estime le délai d'un nouveau client en extrapolant l'historique des attentes des clients actuellement dans la file d'attente, en plus du délai du dernier qui a commencé le service, et en prenant une moyenne pondérée. Le second retourne une moyenne pondérée des délais des anciens clients de la même classe qui ont trouvé la même longueur de file d'attente quand ils sont arrivés. Ensuite, nous nous intéressons à la modélisation des durées de service dans les centres d'appels. En général, les modèles de file d'attente d'Erlang standard sont utilisés pour analyser les opérations dans les centres d'appels. Dans ces modèles, les temps de service des agents sont modélisés comme des variables aléatoires exponentielles indépendantes, identiquement distribuées et de moyenne constante. Plusieurs travaux récents ont montré que la distribution des temps de service est : dépendante du temps, log-normale plutôt qu'exponentielle, et dépend aussi de l'agent. Nous proposons une modélisation plus réaliste des temps de service dans les centres d'appels qui prennent en compte plusieurs propriétés observées dans les données réelles. Nos modèles prennent en compte: l'hétérogénéité des agents, la dépendance du temps, les corrélations sérielles entre les temps de service d'un agent pour le même type d'appel, et les corrélations croisées entre plusieurs types d'appels servis par le même agent. Nous avons montré que ces modèles prédisent les moyennes des temps de service des agents mieux que les modèles de références considérés. Par la suite, nous montrons par la simulation que ces modèles plus réalistes conduisent à des prédictions des performances du système significativement différentes de celles des modèles de références, et les décisions que pourraient prendre le gestionnaire en observant ces données peuvent mener à des économies de coûts importants dans la pratique. / In this thesis, we begin with the study of delay prediction of customers in multiskill call centers. Waiting time has an important impact on the quality of service experienced by customers. Delay announcement can reduce customer uncertainty about its delay time. It also can increase customer satisfaction and reduce the number of abandonments. This requires having a good delay predictor. Unfortunately existing predictors are not adapted for multiskill call centers. We propose three types of predictors that use machine learning: the first uses regression cubic splines, the second employs artificial neural networks, and the latter uses the stochastic kriging. The predictors take as inputs the delay of the last customer of the same type to enter service, the arrival period of the new customer, the staffing of agents groups, the queue length of the same type, and the queue lengths of types served by the same agents. These predictors work well for multiskill call centers, but one drawback is that they have a large number of parameters that must be learned in advance during the training phase that requires a large amount of data and computional time. We also propose two new delay predictors that are very simple to implement, require little optimization effort, do not need any data, and are applicable in multiskill call centers. They are based on the wait times of previous customers of the same class. The first one estimates the delay of a new customer by extrapolating the wait history of customers currently in queue, plus the delay of last one that started service, and taking a weighted average. The second one takes a weighted average of the delays of the past customers of the same class that have found the same queue length when they arrived. Next in this thesis, we are also interested in modelling service time in call centers. In general, the standard Erlang queueing models are used to analyze call centers operations. In these models, agent service times are modelled as independent and identically distributed exponential random variables with a constant mean. Several recent studies have shown that the distribution of service time is: time-dependent, lognormal rather than exponential, and distinct by agent. We propose a more realistic modelling of service times in call centers that takes into account multiple properties observed in real life data. Our models take into account: the heterogeneity of agents, the time dependence, serial correlation between service time of an agent for the same call type, and the cross-correlations between several call types served by the same agent. We show that these models predict agent average service time better than the considered benchmark models. Thereafter, we show by simulation that these more realistic models lead to system performance predictions significantly different from those of the benchmark models, and decisions that manager could take by observing this data can lead to important cost savings in practice.