<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>

We are witnessing a proliferation of massive visual data. Visual content is arguably the fastest growing data on the web. Photo-sharing websites like Flickr and Facebook now host more than 6 and 90 billion photos, respectively. Unfortunately, scaling existing computer vision algorithms to large datasets leaves researchers repeatedly solving the same algorithmic and infrastructural problems. Designing and implementing efficient and provably correct computer vision algorithms is extremely challenging. Researchers must repeatedly solve the same low-level problems: building and maintaining a cluster of machines, formulating each component of the computer vision pipeline, designing new deep learning layers, writing custom hardware wrappers, etc. This thesis introduces CloudCV, an ambitious system that contain algorithms for end-to-end processing of visual content. The goal of the project is to democratize computer vision; one should not have to be a computer vision, big data and deep learning expert to have access to state-of-the-art distributed computer vision algorithms. We provide researchers, students and developers access to state-of-art distributed computer vision and deep learning algorithms as a cloud service through web interface and APIs. / Master of Science

Next-generation sequencing (NGS) technologies have made it possible to rapidly sequence the human genome, heralding a new era of health-care innovations based on personalized genetic information. However, these NGS technologies generate data at a rate that far outstrips Moore\'s Law. As a consequence, analyzing this exponentially increasing data deluge requires enormous computational and storage resources, resources that many life science institutions do not have access to. As such, cloud computing has emerged as an obvious, but still nascent, solution. This thesis intends to investigate and design an efficient framework for running and managing large-scale data-intensive scientific applications in the cloud. Based on the learning from our parallel implementation of a genome analysis pipeline in the cloud, we aim to provide a framework for users to run such data-intensive scientific workflows using a hybrid setup of client and cloud resources. We first present SeqInCloud, our highly scalable parallel implementation of a popular genetic variant pipeline called genome analysis toolkit (GATK), on the Windows Azure HDInsight cloud platform. Together with a parallel implementation of GATK on Hadoop, we evaluate the potential of using cloud computing for large-scale DNA analysis and present a detailed study on efficiently utilizing cloud resources for running data-intensive, life-science applications. Based on our experience from running SeqInCloud on Azure, we present CloudFlow, a feature rich workflow manager for running MapReduce-based bioinformatic pipelines utilizing both client and cloud resources. CloudFlow, built on the top of an existing MapReduce-based workflow manager called Cloudgene, provides unique features that are not offered by existing MapReduce-based workflow managers, such as enabling simultaneous use of client and cloud resources, automatic data-dependency handling between client and cloud resources, and the flexibility of implementing user-defined plugins for data transformations. In-general, we believe that our work attempts to increase the adoption of cloud resources for running data-intensive scientific workloads. / Master of Science

Seamless interconnection of smart mobile devices and cloud services is a key goal in modern mobile computing. Mobile Cloud Computing is the holistic integration of contextually-rich mobile devices with computationally-powerful cloud services to create high value products for end users, such as Apple's Siri and Google's Google Now product. This coupling has enabled new paradigms and fields of research, such as crowdsourced data collection, and has helped spur substantial changes in research fields such as vehicular ad hoc networking. However, the growth of Mobile Cloud Computing has resulted in a number of new challenges, such as testing large-scale Mobile Cloud Computing systems, and increased the importance of established challenges, such as ensuring that a user's privacy is not compromised when interacting with a location-aware service. Moreover, the concurrent development of the Infrastructure as a Service paradigm has created inefficiency in how Mobile Cloud Computing systems are executed on cloud platforms. To address these gaps in the existing research, this dissertation presents a number of software and algorithmic solutions to 1) preserve user locational privacy, 2) improve the speed and effectiveness of deploying and executing Mobile Cloud Computing systems on modern cloud infrastructure, and 3) enable large-scale research on Mobile Cloud Computing systems without requiring substantial domain expertise. / Ph. D.

The increased usage of cloud computing has transformed modern information technology by providing organisations with a scalable, flexible, and cost-effective alternative to the traditional on-premise service model. Both service models have their own set of advantages and disadvantages. One key aspect both service models have in common is the importance of keeping private data secure. There is an ongoing debate on whether cloud computing is safe enough to store private data. This thesis will help organisations understand the security considerations of the different service models. This will be accomplished through a case study researching the different authentication models of both service models and an experiment to gain further insights. The case study and experiment will conclude with a heuristic that organisations can use when picking an authentication model. The main conclusion of this thesis is that we consider the cloud computing service model less secure than the on-premise Windows service model. We also concluded that we consider an LDAP on-premise Windows authentication model and the Azure authentication model to have a higher chance of being less secure than the other authentication models researched in this thesis.

Yes / The governments, nowadays, throughout the world are increasingly becoming dependent on public opinion regarding the framing and implementation of certain policies for the welfare of the general public. The role of social media is vital to this emerging trend. Traditionally, lack of public participation in various policy making decision used to be a major cause of concern particularly when formulating and evaluating such policies. However, the exponential rise in usage of social media platforms by general public has given the government a wider insight to overcome this long pending dilemma. Cloud-based e-governance is currently being realized due to IT infrastructure availability along with mindset changes of government advisors towards realizing the various policies in a best possible manner. This paper presents a pragmatic approach that combines the capabilities of both cloud computing and social media analytics towards efficient monitoring and controlling of governmental policies through public involvement. The proposed system has provided us some encouraging results, when tested for Goods and Services Tax (GST) implementation by Indian government and established that it can be successfully implemented for efficient policy making and implementation.

No / Augmented reality (AR) within the context of higher education is an approach to engage students with experiential learning by utilising AR technology. This paper discusses the process undertaken by a teacher in higher education in designing and implementing cloud-based AR lesson for the students. The methodology engaged was case study at one institution of higher learning in Malaysia. The AR teaching process involves six stages, beginning with the selection of the course, followed by selection of the topic, designing of the AR teaching plan and the implementation of the AR lesson. Upon completion of the implementation of the AR lesson, the teacher and students would provide reflection of their experiences. The process concludes by the improvement of the AR teaching plan by the teacher. The study found that cloud based has indeed disrupted higher education in terms of providing richer learning experiences to the students, as well as enhanced teaching practices for the teachers. Hopefully, this paper would provide insights into the practices of AR teaching and learning approach for teachers in general, and within the context of higher education in particular. It is also intended that the six-steps process outlined in this paper becomes a reference and be duplicated by teachers at large who might be interested to design and implement AR lessons for their own courses.

Yes / Next-generation cellular systems like fth generation (5G) is are expected to experience tremendous tra c growth. To accommodate such tra c demand, there is a need to increase the network capacity that eventually requires the deployment of more base stations (BSs). Nevertheless, BSs are very expensive and consume a lot of energy. With growing complexity of signal processing, baseband units are now consuming a signi cant amount of energy. As a result, cloud radio access networks (C-RAN) have been proposed as anenergy e cient (EE) architecture that leverages cloud computing technology where baseband processing is performed in the cloud. This paper proposes an energy reduction technique based on baseband workload consolidation using virtualized general purpose processors (GPPs) in the cloud. The rationale for the cloud based workload consolidation technique model is to switch o idle baseband units (BBUs) to reduce the overall network energy consumption. The power consumption model for C-RAN is also formulated with considering radio side, fronthaul and BS cloud power consumption. Simulation results demonstrate that the proposed scheme achieves an enhanced energy performance compared to the existing distributed long term evolution (LTE) RAN system. The proposed scheme saves up to 80% of energy during low tra c periods and 12% during peak tra c periods compared to baseline LTE system. Moreover, the proposed scheme saves 38% of energy compared to the baseline system on a daily average.

No / Conventional Call Admission Control (CAC) schemes are based on stand-alone Radio Access Networks (RAN) Base Station (BS) architectures which have their independent and fixed spectral and computing resources, which are not shared with other BSs to address their varied traffic needs, causing poor resource utilization, and high call blocking and dropping probabilities. It is envisaged that in future communication systems like 5G, Cloud RAN (C-RAN) will be adopted in order to share this spectrum and computing resources between BSs in order to further improve the Quality of Service (QoS) and network utilization. In this paper, an intelligent Elastic CAC scheme using Fuzzy Logic in C-RAN is proposed. In the proposed scheme, the BS resources are consolidated to the cloud using virtualization technology and dynamically provisioned using the elasticity concept of cloud computing in accordance to traffic demands. Simulations shows that the proposed CAC algorithm has high call acceptance rate compared to conventional CAC.

Yes / By 2020, next generation (5G) cellular networks are expected to support a 1000 fold traffic increase. To meet such traffic demands, Base Station (BS) densification through small cells are deployed. However, BSs are costly and consume over half of the cellular network energy. Meanwhile, Cloud Radio Access Networks (C-RAN) has been proposed as an energy efficient architecture that leverage cloud computing technology where baseband processing is performed in the cloud. With such an arrangement, more energy gains can be acquired through statistical multiplexing by reducing the number of BBUs used. This paper proposes a green Intelligent Traffic and Resource Elastic Energy (iTREE) scheme for C-RAN. In iTREE, BBUs are reduced by matching the right amount of baseband processing with traffic load. This is a bin packing problem where items (BS aggregate traffic) are to be packed into bins (BBUs) such that the number of bins used are minimized. Idle BBUs can then be switched off to save energy. Simulation results show that iTREE can reduce BBUs by up to 97% during off peak and 66% at peak times with RAN power reductions of up to 27% and 18% respectively compared with conventional deployments.