Confrontées à l'augmentation des coûts de mise en place et de maintenance des systèmes informatiques, les entreprises se tournent vers des solutions d'externalisation telles que le Cloud Computing. Le Cloud se basent sur la virtualisation comme principale technologie permettant la mutualisation. L'utilisation de la virtualisation apporte de nombreux défis donc les principaux portent sur les performances des applications dans les machines virtuelles (VM) et la prévisibilité de ces performances. Dans un système virtualisé, les ressources matérielles sont partagées entre toutes les VMs du système. Dans le cas du CPU, c'est l'ordonnanceur de l'hyperviseur qui se charge de le partager entre tous les processeurs virtuels (vCPU) des VMs. L'hyperviseur réalise une allocation à temps partagé du CPU entre tous les vCPUs des VMs. Chaque vCPU a accès au CPU périodiquement. Ainsi, les vCPUs des VMs n'ont pas accès de façon continue au CPU, mais plutôt discontinue. Cette discontinuité est à l'origine de nombreux problèmes sur des mécanismes tels que la gestion d'interruption et les mécanismes de synchronisation de bas niveau dans les OS invités. Dans cette thèse, nous proposons deux contributions pour répondre à ces problèmes dans la virtualisation. La première est un nouvel ordonnanceur de l'hyperviseur qui adapte dynamiquement la valeur du quantum dans l'hyperviseur en fonction du type des applications dans les VMs sur une plate-forme multi-coeurs. La seconde contribution est une nouvelle primitive de synchronisation (nommée I-Spinlock) dans l'OS invité. Dans un Cloud fournissant un service du type IaaS, la VM est l'unité d'allocation. Le fournisseur établit un catalogue des types de VMs présentant les différentes quantités de ressources qui sont allouées à la VM vis-à-vis des différents périphériques. Ces ressources allouées à la VM correspondent à un contrat sur une qualité de service négocié par le client auprès du fournisseur. L'imprévisibilité des performances est la conséquence de l'incapacité du fournisseur à garantir cette qualité de service. Deux principales causes sont à l'origine de ce problème dans le Cloud: (i) un mauvais partage des ressources entre les différentes VMs et (ii) l'hétérogénéité des infrastructures dans les centres d'hébergement. Dans cette thèse, nous proposons deux contributions pour répondre au problème d'imprévisibilité des performances. La première contribution s'intéresse au partage de la ressource logicielle responsable de la gestion des pilotes, et propose une approche de facturation du temps CPU utilisé par cette couche logiciel aux VMs. La deuxième contribution s'intéresse à l'allocation du CPU dans les Clouds hétérogènes. Dans cette contribution, nous proposons une approche d'allocation permettant de garantir la capacité de calcul allouée à une VM quelle que soit l'hétérogénéité des CPUs dans l'infrastructure. / As a reaction to the increasing costs of setting up and maintaining IT systems, companies are turning to solutions such as Cloud Computing. Cloud computing is based on virtualization as the main technology for mutualisation. The use of virtualization brings many challenges. The main ones concern the performance of the applications in the virtual machines (VM) and the predictability of these performances. In a virtualized system, hardware resources are shared among all VMs in the system. In the case of the CPU, it is the scheduler of the hypervisor that is in charge of sharing the CPU among all the virtual processors (vCPU) of the VMs. The hypervisor uses a time-sharing approach to allocate the CPU. Each vCPU has access to the CPU periodically. Thus, the vCPU of the VMs do not have continuous access to the CPU, but rather discontinuous. This discontinuity is causing many problems on mechanisms such as interuption handling and low-level synchronization mechanisms in guest OSs. In this thesis, we propose two contributions to address these problems in virtualization. The first is a new hypervisor scheduler that dynamically adapts the quantum value in the hypervisor according to the type of applications in the VMs on a multi-core platform. The second contribution is a new synchronization primitive (named I-Spinlock) in the guest OS. In a cloud providing a service of the IaaS type, the VM is the allocation unit. The provider establishes a catalogue presenting the different quantities of resources that are allocated to the VM regarding various devices. These resources allocated to the VM correspond to a contract on a quality of service negotiated by the customer with the provider. The unpredictability of performance is the consequence of the incapability of the provider to guarantee this quality of service. There are two main causes of this problem in the Cloud: (i) poor resource sharing between different VMs and (ii) heterogeneity of infrastructure in hosting centers. In this thesis, we propose two contributions to answer the problem of performance unpredictability. The first contribution focuses on the sharing of the software resource responsible for managing the drivers, and proposes to bill the CPU time used by this software layer to VMs. The second contribution focuses on the allocation of the CPU in heterogeneous clouds. In this contribution, we propose an allocation approach to guarantee the computing capacity allocated to a VM regardless of the heterogeneity of the CPUs in the infrastructure.

Allocation and reservation of resources, such as CPU cycles and I/O bandwidth of multimedia servers and link bandwidth in the network, is essential to ensure Quality of Service (QoS) of multimedia services delivered over the Internet. We propose a Distributed Multimedia Server System (DMSS) configured out of a collection of networked multimedia servers where multimedia data are partitioned and replicated among the servers. We also introduce Utility Model-Distributed (UM-D), the distributed version of the Utility Model, for admission control and QoS adaptation of multimedia sessions to maximize revenue from multimedia services for the DMSS. Two control architectures, a centralized and a distributed, have been proposed to solve the admission control problem formalized by the UM-D. In the centralized broker architecture, the admission control in a DMSS can be mapped to the Multidimensional Multiple-choice Knapsack Problem (MMKP), a variant of the classical 0–1 Knapsack Problem. An exact solution of MMKP, an NP-hard problem, is not applicable for the on line admission control problem in the DMSS. We therefore developed three new heuristics, M-HEU, I-HEU and C-HEU for solving the MMKP for on-line real-time admission control and QoS adaptation. We present a qualitative analysis of the performance of these heuristics to solve admission control problems based on the worst-case complexity analysis and the experimental results from different sized data sets. The fully distributed admission control problem in a DMSS, on the other hand, maps to the Multidimensional Multiple-choice Multi Knapsack Problem (MMMKP), a new variant of the Knapsack Problem. We have developed D-HEU and A-HEU, two new distributed heuristics to solve the MMMKP. D-HEU requires a large number of messages and it is not suitable for a on line admission controller. A-HEU finds the solution with fewer messages but achieves less optimality than D-HEU. We have applied the admission control strategy described in the UM-D to the set of Media Server Farms providing streaming videos to users. The performance of different heuristics in the broker has been discussed using the simulation results. We have also shown application of UM-D to Distributed SLA (Service Level Agreement) Controllers in Enterprise Networks. Simulation results and qualitative comparison of different heuristics are also provided. / Graduate

Orientadores: Mauricio Ferreira Magalhães, Juan Manuel Adan-Coelho / Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Engenharia Eletrica e de Computação / Made available in DSpace on 2018-08-14T23:24:48Z (GMT). No. of bitstreams: 1 MeloJunior_Alencarde_D.pdf: 1273545 bytes, checksum: afd8be737ed5bdf8e7661aa96cd2ca60 (MD5) Previous issue date: 2009 / Resumo: Muitas aplicações que se adaptam aos recursos disponíveis na rede têm se tornado populares ultimamente na Internet, mostrando que níveis absolutos de Qualidade de Serviço (QoS) não são essenciais para a maioria das aplicações. O Modelo de Diferenciação Proporcional (MDP) constitui-se atualmente em uma das principais propostas de serviço para a Internet de Próxima Geração e pode acrescentar mais uma possibilidade ao espaço de dimensões disponíveis para a adaptação das aplicações. Esta tese investiga o MDP, situando-o diante de outras propostas de serviços existentes e propõe os algoritmos para escalonamento de pacotes Window Based Waiting-Time Priority (WBWTP) e Window Based Waiting-Time Priority Plus (WBWTP+), os quais possibilitam atingir os objetivos do MDP com maior precisão do que os tradicionais algoritmos Waiting-Time Priority (WTP) e Proportional Average Delay (PAD). É também apresentada uma métrica de QoS para avaliar o desempenho dos algoritmos, que fornece uma medida de desempenho agregada para todas as classes de serviço ao longo de um intervalo de monitoração, facilitando a compreensão da dinâmica do MDP. / Abstract: Many applications that adapt to available resources have became popular lately in the Internet, suggesting that absolute levels of Quality of Services (QoS) are not essential for most applications. The Proportional Differentiation Model (PDM) constitutes currently one of the main proposals for the Next Generation Internet and adds one more possibility for application adaptation. This dissertation investigates the PDM, situating it along other existing service proposals and presents the Window Based Waiting-Time Priority (WBWTP) and Window Based Waiting-Time Priority Plus (WBWTP+) packet scheduling algorithms, which enable to achieve the PDM objectives with higher accuracy than the traditional Waiting-Time Priority (WTP) and Proportional Average Delay (PAD) algorithms. The performance of the algorithms is evaluated using a proposed QoS metric that supplies a single performance measurement for all service classes along a monitoring interval. / Doutorado / Engenharia de Computação / Doutor em Engenharia Elétrica

Dr. Renalde Huysamen

A traffic conditioner is an element of the Differentiated Services architecture. This architecture is used to regulate quality of service in computer networks. Five traffic conditioners were selected for the study. These include the token bucket marker and four conditioners described in RFCs. The contribution of this dissertation is two-fold. Firstly, it presents process algebra models of the five identified traffic conditioners. These models provide succinct descriptions of the conditioners thereby highlighting essential features. The models are, however, not intended for model checking purposes, but rather serve as a convenient pedagogical device. The second and main contribution of the dissertation is a simulation study to investigate the relative performance of the five traffic conditioners across a range of simulated scenarios in which traffic patterns and subscription levels are varied in a fixed network topology. Two performance measures—TargetRatio and GreenRatio—are defined, justified, and used to compare the traffic conditioners. The GreenRatio measure was found to be more discriminating than the TargetRatio measure. A variant of the GreenRatio measure was used to further illuminate the differences between conditioners. The simulation results suggest that the performance of the conditioners are sensitive to parameter values such as token bucket size and that bursty traffic patterns are particularly sensitive to these parameters. Under such bursty conditions, these parameters should be chosen with care. / Dissertation (MSc (Computer Science))--University of Pretoria, 2007. / Computer Science / unrestricted

In future, Mobile Ad Hoc Networks (MANETs) may provide access to services in the Internet. MANETs should therefore support diverse applications and data types. This introduces a need for quality of service (QoS), a process of discriminating different data types to provide them with an appropriate level of service. However, QoS can be affected by nodes performing packet forwarding attacks. A critical analysis of the related literature shows that research into QoS and security has typically proceeded independently. However, QoS and security should be considered together as attacks may adversely affect QoS. A simulation study demonstrates this by investigating two single-path packet forwarding approaches under a range of conditions. The study shows that using single-path packet forwarding in the presence of attackers is generally insufficient to support QoS.Based on this background research, a novel 2-Dimensional Adaptation ARChitecture (2-DAARC) and a Priority-based Multi-path Type Selection (PMTS) algorithm are proposed. 2-DAARC integrates two modes of adaptation. The single-path adaptation (SPA) mode uses adaptive bandwidth reservations over a single path for QoS in the presence of node mobility. The multi-path adaptation (MPA) mode uses duplicated data packet transmissions over multiple paths for QoS in the presence of packet forwarding attackers. Adaptation occurs within and between modes to optimize priority packet forwarding in the dynamic MANET environment. The MPA mode uses the PMTS algorithm to select a secondary path which is maximally-disjoint with the primary path. This aims to select a path which may enhance reliability whilst keeping the costs of path selection low. Simulating 2-DAARC shows that under light loads it achieves better QoS than related work, but with a higher control packet overhead. Simulating PMTS shows that under light loads it achieves packet deliveries which are at best as good as a related approach, with lower end-to-end delays and control packet overhead. A novel Congestion and ATtack (CAT) detection mechanism is proposed to improve the performance of 2-DAARC in heavily loaded networks. CAT detection differentiates the causes of packet loss so that adaptation can be better tailored to the network conditions. Without CAT detection, 2-DAARC uses the MPA mode in congested conditions, and this worsens QoS. Simulating 2-DAARC with CAT detection shows that it generally achieves packet deliveries which are greater than or similar to, and end-to-end delays which are less than or similar to related work, and it does so with a lower control packet overhead.

Includes bibliographical references (leaves 80-84). / Recent times has seen a tremendous increase in the deployment and use of 802.11 Wireless Local Area Networks (WLANs). These networks are easy to deploy and maintain, while providing reasonably high data rates at a low cost. In the paradigm of Next-Generation-Networks (NGNs), WLANs can be seen as an important access network technology to support IP multimedia services. However a traditional WLAN does not provide Quality of Service (QoS) support since it was originally designed for best effort operation. The IEEE 802. 11e standard was introduced to overcome the lack of QoS support for the legacy IEEE 802 .11 WLANs. It enhances the Media Access Control (MAC) layer operations to incorporate service differentiation. However, there is a need to prevent overloading of wireless channels, since the QoS experienced by traffic flows is degraded with heavily loaded channels. An admission control scheme for IEEE 802.11e WLANs would be the best solution to limit the amount of multimedia traffic so that channel overloading can be prevented. Some of the work in the literature proposes admission control solutions to protect the QoS of real-time traffic for IEEE 802.11e Enhanced Distributed Channel Access (EDCA). However, these solutions often under-utilize the resources of the wireless channels. A measurement-aided model-based admission control scheme for IEEE 802.11e EDCA WLANs is proposed to provide reasonable bandwidth guarantees to all existing flows. The admission control scheme makes use of bandwidth estimations that allows the bandwidth guarantees of all the flows that are admitted into the network to be protected. The bandwidth estimations are obtained using a developed analytical model of IEEE 802.11e EDCA channels. The admission control scheme also aims to accept the maximum amount of flows that can be accommodated by the network's resources. Through simulations, the performance of the proposed admission control scheme is evaluated using NS-2. Results show that accurate bandwidth estimations can be obtained when comparing the estimated achievable bandwidth to actual simulated bandwidth. The results also validate that the bandwidth needs of all admitted traffic are always satisfied when the admission control scheme is applied. It was also found that the admission control scheme allows the maximum amount of flows to be admitted into the network, according the network's capacity.

To achieve greater performance in computing networks, a setup of critical computing aspects that ensures efficient network operation, needs to be implemented. One of these computing aspects is, Quality of Service (QoS). Its main functionality is to manage traffic queues by means of prioritizing sensitive network traffic. QoS capable networking allows efficient control of traffic especially for network critical data. However, to achieve this in Wireless Sensor Networks (WSN) is a serious challenge, since these technologies have a lot of computing limitations. It is even difficult to manage networking resources with ease in these types of technologies, due to their communication, processing and memory limitations. Even though this is the case with WSNs, they have been largely used in monitoring/detection systems, and by this proving their application importance. Realizing efficient network control requires intelligent methods of network management, especially for sensitive network data. Different network types implement diverse methods to control and administer network traffic as well as effectively manage network resources. As with WSNs, communication traffic and network resource control are mostly performed depending on independently employed mechanisms to deal with networking events occurring on different levels. It is therefore challenging to realize efficient network performance with guaranteed QoS in WSNs, given their computing limitations. Software defined networking (SDN) is advocated as a potential paradigm to improve and evolve WSNs in terms of capacity and application. A means to apply SDN strategies to these compute-limited WSNs, formulates software defined wireless sensor networks (SDWSN). In this work, a resource-aware OpenFlow-based Active Network Management (OF-ANM) QoS scheme that uses SDN strategies is proposed and implemented to apply QoS requirements for managing traffic congestion in WSNs. This scheme uses SDN programmability strategies to apply network QoS requirements and perform traffic load balancing to ensure congestion control in SDWSN. Our experimental results show that the developed scheme is able to provide congestion avoidance within the network. It also allows opportunities to implement flexible QoS requirements based on the system’s traffic state. Moreover, a QoS Path Selection and Resource-associating (Q-PSR) scheme for adaptive load balancing and intelligent resource control for optimal network performance is proposed and implemented. Our experimental results indicate better performance in terms of computation with load balancing and efficient resource alignment for different networking tasks when compared with other competing schemes. / Thesis (PhD)--University of Pretoria, 2019. / National Research Foundation / University of Pretoria / Electrical, Electronic and Computer Engineering / PhD / Unrestricted

The main aim of my master thesis was to get acquainted with the implementation of quality-of-service support methods in UMTS 3G mobile network and the evaluation of these methods in a simulation environment with main interest in the characteristic features of the traffic classes. The impact of the QoS methods available on different service types were verified in the OPNET Modeler simulation environment. I examined mainly the potentials of the model to support the required quality parameters for the connections. The models of typical network applications, available in real UMTS networks, together with two examples of mobile networks, operating voice and video services in combination with standard data services, were defined in this simulation environment. My work also contains the detailed description of configuration parameters fundamental for correct behavior of the simulation models. The simulation also shows the case, when the transport service is rejected due to the lack of network resources. The key parameters of the model are also described and their effect on the service quality with and without QoS control are compared in separate scenarios. Finally, I investigated the effect of mobility of the end stations on the communication process, especially on the duration of a file transmission.

A common trend, in general as well as in the field of combat vehicles, is the rapidly increasing demand for data network capacity and even more in transferred data. To handle this increased demand, different countries with their armed forces and equipment manufacturers evaluate methods to increase the data transmission capacity in combat vehicles. The different types of transmitted data are of different criticality and have different security demands. An easy solution to this is to have separated networks for each type of traffic, but that is quite expensive and uses a lot of hardware. This thesis focuses on a different solution, with a shared network for all types of data transmissions. This is done by evaluating different types of data networks and add-on protocols and then testing the networks practically with varying transmission rates. In the thesis, all the practical testing is done with data networks according to the Ethernet standard, which is the standard evaluated with a throughput that is high enough for the use case. Ethernet as a standard is not suitable for critical data traffic and therefore add-on protocols for Ethernet to optimize the system for critical data traffic are tested. With these optimizations made, Ethernet can be considered more suitable for critical traffic, but this depends entirely on the system requirements.