Cloud Computing är det stora modeordet i IT-världen just nu. Det har blivit mer och mer populärt på senare år men frågor har uppstått om dess prestanda och säkerhet. Hur säkert är det egentligen och är det någon större skillnad i prestanda mellan en lokal server och en molnbaserad server? Detta examensarbete tar upp dessa frågor. En serie prestandatester kombinerat med en litteraturstudie genomfördes för att få fram ett resultatet för detta examensarbete.Denna rapport kan komma att vara till nytta för de som har ett intresse av Cloud Computing men som saknar någon större kunskap om ämnet. Resultaten kan användas som exempel för hur framtida forskning inom Cloud Computing kan genomföras. / Cloud Computing - the big buzz word of the IT world. It has become more and more popular in recent years but questions has arisen about it’s performance and security. How safe is it and is there any real difference in performance between a locally based server and a cloud based server? This thesis will examine these questions. A series of performance tests combined with a literature study were performed to achieve the results of this thesis.This thesis could be of use for those who have an interest in Cloud Computing and do not have much knowledge of it. The results can be used as an example for how future research in Cloud Computing can be done.

No / Over the last few years, cloud computing has become quite popular. It offers Web-based companies the advantage of scalability. However, this scalability adds complexity which makes analysis and predictable performance difficult. There is a growing body of research on load balancing in cloud data centres which studies the problem from the perspective of the cloud provider. Nevertheless, the load balancing of scalable web servers deployed on the cloud has been subjected to less research. This paper introduces a simple queueing model to analyse the performance metrics of web server under varying traffic loads. This assists web server managers to manage their clusters and understand the trade-off between QoS and cost. In this proposed model two thresholds are used to control the scaling process. A discrete-event simulation (DES) is presented and validated via an analytical solution.

Yes / Failure is an increasingly important issue in high performance computing and cloud systems. As large-scale systems continue to grow in scale and complexity, mitigating the impact of failure and providing accurate predictions with sufficient lead time remains a challenging research problem. Traditional existing fault-tolerance strategies such as regular check-pointing and replication are not adequate because of the emerging complexities of high performance computing systems. This necessitates the importance of having an effective as well as proactive failure management approach in place aimed at minimizing the effect of failure within the system. With the advent of machine learning techniques, the ability to learn from past information to predict future pattern of behaviours makes it possible to predict potential system failure more accurately. Thus, in this paper, we explore the predictive abilities of machine learning by applying a number of algorithms to improve the accuracy of failure prediction. We have developed a failure prediction model using time series and machine learning, and performed comparison based tests on the prediction accuracy. The primary algorithms we considered are the Support Vector Machine (SVM), Random Forest(RF), k-Nearest Neighbors (KNN), Classi cation and Regression Trees (CART) and Linear Discriminant Analysis (LDA). Experimental results indicates that the average prediction accuracy of our model using SVM when predicting failure is 90% accurate and effective compared to other algorithms. This f inding implies that our method can effectively predict all possible future system and application failures within the system. / Petroleum Technology Development Fund (PTDF) funding support under the OSS scheme with grant number (PTDF/E/OSS/PHD/MB/651/14)

The southeastern United States is one of the most productive forestry regions in the world, encompassing approximately 100 million ha of forest land, about 87% of which is privately owned. Any alteration in this region's duration or rate of forest recovery has consequential economic and ecological ramifications. Despite the need for forest recovery monitoring in this region, a spatially comprehensive evaluation of forest spectral recovery through time has not yet been conducted. Remote sensing analysis via cloud-computing platforms allows for evaluating southeastern forest recovery at spatiotemporal scales not attainable with traditional methods. Forest productivity is assessed in this study using spectral metrics of southern yellow pine recovery following stand-replacing disturbance. An annual cloudfree (1984-2021) Landsat time series intersecting ten southeastern states was constructed using the Google Earth Engine API. Southern yellow pine stands were detected using the National Land Cover Database (NLCD) evergreen class, and pixels with a rapidly changing spectrotemporal profile, suggesting stand-replacing disturbance, were found using the Landscape Change Monitoring System (LCMS) Fast Loss product. Spectral recovery metrics for 3,654 randomly selected stands in 14 Level 3 EPA Ecoregions were derived from their 38-year time series of Normalized Burn Ratio (NBR) values using the Detecting Breakpoints and Estimating Segments in Trend (DBEST) change detection algorithm. Recovery metrics characterizing the rate (NBRregrowth), duration (Y2R), and magnitude (K-shift) of recovery from stand-replacing disturbances occurring between 1989 and 2011 were evaluated to identify long-term and wide-scale changes in forest recovery using linear regression and spatial statistics respectively. Sampled stands typically recover 35% higher in NBR than pre-disturbance and, on average, spectrally recover within seven years of disturbance. Recovery rate is shown to be increasing over time; temporal slope estimates for NBRregrowth suggest a 33% increase in early recovery rate between 1984 and 2011. Similarly, recovery duration measured with Y2R decreased by 43% during the study period with significant spatial variation. Results suggest that the magnitude of change in stand condition between rotations has decreased by 21% during the study period, has substantial regional divisions in high and low magnitude recovery between coastal and inland stands, and low NBR value sites have the most potential to increase their NBR value. Observed spatiotemporal patterns of spectral recovery suggest that changes in management interventions, atmospheric CO2, and climate over time have changed regional productivity. Results from this study will aid the understanding of changing productivity in southern yellow pine and will inform the management, monitoring, and modeling of this ecologically and economically important forest ecosystem. / Master of Science / The Southeast United States contains approximately 100 million hectares of forest land and is one of the world's most productive regions for commercial forestry. Forest managers and those who model the effects of different types of forest land on the changing climate need up-to-date information about how productive these forests are at removing carbon and producing wood and how that productivity differs across space and time. In this study, we evaluate the productivity of southern yellow pine stands by measuring stand recovery attributes from a disturbance that removes the majority or all of the trees in the stand. This is accomplished by locating 3,654 of randomly selected disturbed pine stands through ten southeastern states using freely available national data products derived from Landsat satellite imagery, namely a combination of the National Land Cover Database (NLCD) and the Landscape Change Monitoring System (LCMS), which provide information about the type of forest, and the year and severity of disturbance respectively. Annual Landsat satellite imagery from 1984 to 2021 is used to create a series of values over time for each stand representing the stand condition each year using an index called the Normalized Burn Ratio (NBR). A change detection algorithm called DBEST is applied to each stands NBR values to find the timing of disturbance and recovery, which is used to create three metrics characterizing the rate (NBRregrowth), duration (Y2R), and magnitude (K-shift) of recovery. We evaluated how these metrics change through time using linear regression and how they differ across space using regression residuals and spatial statistics. Across the region, stands typically increase in recovery rate, decrease in recovery duration, and decrease in recovery magnitude. On average, stands recover within seven years of disturbance and to a higher NBR value than pre-disturbance. However, there is significant spatial variation in this metric throughout the Southeast. The results indicate that stands with a lower vegetation condition, measured with NBR, before the disturbance had the most significant gain in stand condition after recovery, and stands with initially higher vegetation condition did not increase as much after recovery.

In today’s world, IoT systems are more and more overwhelming. All electronic devices are becoming connected. From lamps and refrigerators in smart homes, smoke detectors and cameras in monitoring systems, to scales and thermometers in healthcare systems, until phones, cars and watches in smart cities. All these connected devices generate a huge amount of data collected from the environment. To take advantage of these data, a processing phase is needed in order to extract useful information, allowing the best management of the system. Since most objects in IoT systems are resource limited, the processing step, usually performed by an artificial intelligence model, is offloaded to a more powerful machine such as the cloud server in order to benefit from its high storage and processing capacities. However, the cloud server is geographically remote from the connected device, which leads to a long communication delay and harms the effectiveness of the system. Moreover, due to the incredibly increasing number of IoT devices and therefore offloading operations, the load on the network has increased significantly. In order to benefit from the advantages of cloud based AIoT systems, we seek to minimize its shortcomings. In this thesis, we design a distributed architecture that allows combining these three domains while reducing latency and bandwidth consumption as well as the IoT device’s energy and resource consumption. Experiments conducted on different cloud based AIoT systems showed that the designed architecture is capable of reducing up to 80% of the transmitted data. / En el mundo actual, los sistemas de IoT (Internet de las cosas) son cada vez más abrumadores. Todos los dispositivos electrónicos se están conectando entre sí. Desde lámparas y refrigeradores en hogares inteligentes, detectores de humo y cámaras para sistemas de monitoreo, hasta básculas y termómetros para sistemas de atención médica, pasando por teléfonos, automóviles y relojes en ciudades inteligentes. Todos estos dispositivos conectados generan una enorme cantidad de datos recopilados del entorno. Para aprovechar estos datos, es necesario un proceso de análisis para extraer información útil que permita una gestión óptima del sistema. Dado que la mayoría de los objetos en los sistemas de IoT tienen recursos limitados, la etapa de procesamiento, generalmente realizada por un modelo de inteligencia artificial, se traslada a una máquina más potente, como el servidor en la nube, para beneficiarse de su alta capacidad de almacenamiento y procesamiento. Sin embargo, el servidor en la nube está geográficamente alejado del dispositivo conectado, lo que conduce a una larga demora en la comunicación y perjudica la eficacia del sistema. Además, debido al increíble aumento en el número de dispositivos de IoT y, por lo tanto, de las operaciones de transferencia de datos, la carga en la red ha aumentado significativamente. Con el fin de aprovechar las ventajas de los sistemas de AIoT (Inteligencia Artificial en el IoT) basados en la nube, buscamos minimizar sus desventajas. En esta tesis, hemos diseñado una arquitectura distribuida que permite combinar estos tres dominios al tiempo que reduce la latencia y el consumo de ancho de banda, así como el consumo de energía y recursos del dispositivo IoT. Los experimentos realizados en diferentes sistemas de AIoT basados en la nube mostraron que la arquitectura diseñada es capaz de reducir hasta un 80% de los datos transmitidos.

Forecasting has its application in various domains as the decision-makers are provided with a more predictable and reliable estimate of events that are yet to occur. Typically, a user would invest in licensed software or subscribe to a monthly or yearly plan in order to make such forecasts. The framework presented here differs from conventional software in forecasting, as it allows any interested party to use the proposed services on a pay-per-use basis so that they can avoid investing heavily in the required infrastructure. The Framework-as-a-Service (FaaS) presented here uses Windows Communication Foundation (WCF) to implement Service-Oriented Architecture (SOA). For forecasting, collection of data, its analysis and forecasting responsibilities lies with users, who have to put together other tools or software in order to produce a forecast. FaaS offers each of these responsibilities as a service, namely, External Data Collection Framework (EDCF), Internal Data Retrieval Framework (IDRF) and Forecast Generation Framework (FGF). FaaS Controller, being a composite service based on the above three, is responsible for coordinating activities between them. These services are accessible through Economic Endpoint (EE) or Technical Endpoint (TE) that can be used by a remote client in order to obtain cost or perform a forecast, respectively. The use of Cloud Computing makes these services available over the network to be used as software to forecast energy for solar or wind resources. These services can also be used as a platform to create new services by merging existing functionality with new service features for forecasting. Eventually, this can lead to faster development of newer services where a user can choose which services to use and pay for, presenting the use of FaaS as Platform-as-a-Service (PaaS) in forecasting. / Master of Science

En los últimos años y con el auge las tecnologías de virtualización y de las infraestructuras cloud, se abre un nuevo abanico de posibilidades para acceso de recursos de cómputo para el ámbito científico. Estas tecnologías permiten "acceso ubicuo, adaptado y bajo demanda en red a un conjunto compartido de recursos de computación". Estas tecnologías permiten que el acceso a grandes cantidades de recursos virtualizados sea mucho más sencillo para el científico. Si bien la adaptación de aplicaciones a un entorno distribuido sigue requiriendo de una experiencia importante, es posible utilizar de forma eficiente software adaptado a sistemas de colas e incluso computación paralela de memoria distribuida. A pesar de todo, en la actualidad existen diferentes proveedores cloud, diferente software para el despliegue de plataformas cloud, diferentes gestores de máquinas virtuales, y otros componentes que complican el acceso de forma sencilla y homogénea. Por tanto el objetivo principal de esta tesis es la de proporcionar a la comunidad científica el acceso a las tecnologías de virtualización y cloud de manera sencilla. De tal manera que sea muy sencillo el despliegue y gestión de sus infraestructuras virtuales, para que los investigadores solo tengan que centrarse en las tareas propias de su aplicación. Una plataforma Cloud para investigación debe contemplar todos los aspectos necesarios para creación y gestión de las infraestructuras, partiendo de que el investigador debe poder expresar sus requerimientos, tanto hardware como software, sobre los recursos que va a necesitar para la ejecución de su aplicación. En base a los requerimientos definidos por el usuario el sistema debe crear la infraestructura del usuario, teniendo en cuenta aspectos como la selección de despliegues cloud, de imágenes de máquinas virtuales, procesos de contextualización, etc. El sistema también debe permitir que el usuario modifique la cantidad de recursos (elasticidad horizontal) así como las características de los mismos (elasticidad vertical). Por último la plataforma debe proporcionar interfaces tanto a nivel de usuario, mediante aplicaciones de comandos o interfaces gráficas, como a nivel programático para que capas de mayor nivel puedan hacer uso de la funcionalidad mediante un API. La tesis pretende tanto avanzar en las especificaciones y arquitecturas software como desarrollar y testear un prototipo. / Caballer Fernández, M. (2014). Gestión de infraestructuras virtuales configuradas dinámicamente [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/37376 / Premios Extraordinarios de tesis doctorales

Our daily lives are increasingly reliant on Internet-based services, which in turn are increasingly dependent upon a small number of cloud providers to support their functionality. Internet service consolidation aggrandizes the select few cloud providers who are consolidation’s beneficiaries. As cloud providers’ networks – primarily those of Amazon, Google, IBM, and Microsoft – become more vital to the Internet’s operation and functionality, it is paramount that we understand the different factors that affect their performance and how they are affecting the greater Internet. It is also imperative that researchers are able to conduct independent measurements and tests regarding these topics without requiring collaboration with, or special access from, the cloud providers. Due to proprietary implementations, the results of collaborative studies or experiments lack transparency and are difficult, if not impossible, for independent researchers to repeat. In this dissertation we seek to understand how the cloud providers’ networks impact Internet user performance versus using the public Internet and how they are affecting the Internet’s topological structure. To achieve this, first we enable future research to test how various net- work component interactions affect performance by creating a testbed that approximates the cloud providers’ network environment. The testbed provides researchers with the same control and fidelity as the operators of a complex network, such as the cloud providers. No prior work provided such a platform for Internet networking research. Second, we examine how the cloud providers’ private Wide Area Networks (WANs) impact user performance. We do this by leveraging new services provided by Google and Amazon which enables us to isolate the latency differences between using their private WANs versus us- ing the public Internet. We conduct the first comprehensive study that quantifies the performance difference, sourced from networks we estimate to represent 91% of the Internet’s user population. We then examine several case studies in specific regions where one of the two networks provides significant performance benefits and examine why these outliers occur. This body of work shows that using only measurements and analyses that do not require special access or permissions from the cloud providers, we can discern that the cloud providers’ private WANs deliver modest performance improvements over the public Internet. Third, we examine the impact the cloud providers’ networks have on the Internet and its topology. We start by investigating and quantifying to what extent increased connectivity between the cloud providers and other networks allows the cloud providers to bypass the hierarchical Inter- net. We then analyze the geographic deployment strategy of the cloud providers, compare it against the Tier-1 and Tier-2 Internet Service Providers (ISPs), and examine the population coverage of each based on geographic proximity to network resources. We show that the cloud providers’ extensive peering footprints provide the potential for them to reach the majority of networks while bypassing the hierarchical Internet.

<p dir="ltr">This thesis addresses critical optimization challenges in large-scale, heterogeneous data centers: logical cluster formation for virtual machine placement and physical rack movement for efficient infrastructure management. As data centers grow in size and complexity, these systems face rising demands to minimize costs related to fault tolerance, reformation, and resource constraints while adapting to diverse hardware and operational requirements. </p><p dir="ltr">The first part focuses on logical cluster formation, where capacity guarantees must be maintained across millions of servers despite ongoing infrastructure events, such as maintenance and failures. Traditional offline methods fall short under these dynamic, large-scale conditions. To address this, a two-tier approach combining deep reinforcement learning (DRL) with mixed-integer linear programming (MILP) enables real-time resource allocation, reducing server relocations and enhancing resilience across complex server environments.</p><p dir="ltr">The second part tackles optimized rack placement in highly heterogeneous settings, where balancing fault tolerance, energy efficiency, and load distribution is essential. Static layouts struggle to accommodate diverse hardware configurations and fluctuating resource needs. This research proposes a scalable, tiered optimization approach using the Leader Reward method and a gradient-based heuristic to handle the computational demands of large-scale rack positioning.</p><p dir="ltr">By integrating DRL and heuristic techniques, this work provides a robust, scalable solution for cost efficiency and operational resilience in managing large, heterogeneous data centers, advancing intelligent data center management for modern cloud infrastructure.</p>

<p dir="ltr">Container orchestration systems, such as Kubernetes, often rely on manual resource allocation to manage resources, which can be inefficient and inflexible due to frequent over-provisioning or underprovisioning. Kubernetes horizontal pod autoscaler (HPA), vertical pod autoscaler (VPA), and Google Kubernetes Engine (GKE) Autopilot are primarily threshold-based, making them reactive rather than proactive since they adjust resources after exceeding utilization thresholds, leading to temporary degradation in quality of service~(QoS). While some solutions utilize calls per minute (CPM) counts for requests to microservices to estimate resource consumption dynamically, they do not fully exploit distributed traces or associated microservices' interdependencies. This thesis hypothesizes that more profound insights into future workload patterns can be gained by exploiting microservices' interaction and the CPM counts for each pair of communicating microservices. This thesis proposes a comprehensive machine learning workflow to assess whether factoring in the interdependencies between microservices results in improved workload prediction. The findings of this study indicate that a long short-term memory (LSTM) model performs well, with average mean absolute error (MAE) and root mean square error (RMSE) values of 7.02 and 10.54, respectively. The highest \(R^2\) score observed was 0.07. This suggests that although incorporating distributed traces and inter-microservice CPM counts provides valuable insights, the models fail to capture the full complexity of workload dynamics. These results highlight the potential for enhancing workload prediction accuracy and underscore the need to refine these methods further to achieve more proactive and efficient resource allocation in container orchestration systems.</p>