Dysarthria is a motor speech disorder resulting in slurred or slow speech that can be difficult to understand. This re- search paper evaluates the effectiveness of various metrics for automatic speech recognition (ASR), such as character error rate (CER), Jaro-Winkler distance, and BERTscore, in assessing performance specifically for dysarthric speech, which is often inadequately measured by the commonly used word error rate (WER). Using the TORGO database, which includes a range of dysarthria severities, we analyze the performance of chosen evaluation metrics with the Whisper and wav2vec 2.0 ASR systems to understand how they reflect the true speech recognition challenges presented by such atypical speech pat- terns. Our findings reveal that Whisper generally outperforms wav2vec 2.0, particularly in sentence utterances, by effectively managing complex speech patterns and maintaining semantic integrity. The analysis highlights that single-word utterances strongly correlate with dysarthria severity, while sentence utterances show a lesser correlation due to the mitigating effect of additional linguistic context.

Containerization with Docker has become the standard for the deployment of software in recent years since it is a lightweight method to isolate applications. However, the selection of a Docker image brings different dependencies, which can introduce energy inefficiencies that are often not documented. Recent work performs a comparative study between different base images and finds certain patterns in which some images tend to be less energy efficient for certain tasks. In this thesis, we propose a methodology to find the reason behind observed energy inefficiencies in a Docker image by using tracing to study how a workload makes use of the dependencies provided by the image. We apply this methodology to test the energy efficiency of Redis when using different base images. This leads to finding and raising awareness of energy inefficiencies in some functions from musl, the C standard library implementation used by Alpine.

Software development often relies on dependencies managed by package managers to simplify the integration of external libraries and frameworks, reducing development time. However, developers sometimes choose to bundle dependencies directly within their software packages. Bundling dependencies means including all necessary third-party frameworks directly within the application's distributable archive, such as a JAR file, to ensure all components are present without needing external installations. This practice, resulting in Uber JARs (or fat JARs), presents both challenges and advantages within the Maven ecosystem. This project examines the prevalence, risks, and impact of Uber JARs by analyzing over 9 million POM files and 12 million JAR artifacts from Maven Central, identifying artifacts with previously undetected vulnerabilities. Notably, 10.48% of the analyzed artifacts, amounting to 915,089, fall under the category of Uber JARs, indicating a significant prevalence within the Maven repository. Central to this work, JarSift detects Uber JARs' contents, including the libraries, their versions, and vulnerabilities. JarSift's accuracy is demonstrated with an F1 score ranging from 0.474 to 0.857, depending on the Uber JAR configuration. Analysis reveals about 17.13% Uber JARs in a small-scale dataset contained undisclosed vulnerabilities, and 0.63% of all libraries in our dataset fully completely matched known vulnerable libraries. These findings highlight the need for better detection and mitigation strategies in the Maven ecosystem and inform developers of potential risks, helping them implement more robust security measures.

The global climate change crisis and the associated phenomenon of global warming have taken center stage in recent years. Greenhouse gas emissions due to electricity generation are a contributor to this problem. Internet Services running in data centers consume enormous amounts of energy and should be optimized to reduce their greenhouse gas emissions.

This thesis explores the possibility of intelligently scheduling resource-intensive batch data-processing jobs to green energy generation hours in the day. Green hours are hours within the day during which the amount of greenhouse gasses emitted per kilowatt-hour (kWh) is lower compared to other hours of the day. There is a variance in the amount of emissions due to the variability of renewable energy generation and grid demand.

The system "S.C.A.L.E. (Scheduler for Carbon-Aware Load Execution)" is proposed. It schedules compute jobs to periods of low-carbon-intense energy generation based on predictions of renewable energy generation and grid demand. The system was evaluated against a simulated data processing pipeline at ING; this pipeline is one of the larger consumers of the ING private cloud. The scheduler aims to reduce greenhouse gas emissions by intelligently predicting task running times and green hours for the next day and optimizing the times at which tasks are processed throughout the day.

Several main conclusions are drawn based on this research:
  1. The accuracy of task load predictions regarding running times is crucial for effective scheduling. The research concludes that, with sufficient historical data, the scheduler can predict task running times with an acceptable margin of error (5-10%).
  2. The research explores the scheduler's ability to predict periods of low carbon intensity and the resulting reduction in carbon emissions by implementing it. The research affirms the scheduler's accuracy in determining low-carbon-intensive energy generation periods and estimates a potential 20% reduction in greenhouse gas emissions.
  3. The potential overhead introduced by implementing a carbon-aware scheduler is addressed. The research identifies that while the scheduling algorithm itself is lightweight, the concurrent processing of tasks introduces overhead. The tipping point, where the overhead outweighs the benefits, varies for each system and should be experimentally determined.

The thesis concludes by emphasizing the significance of implementing a carbon-aware scheduler to reduce the environmental impact of data centers. The proposed scheduler is a promising contribution to sustainable computing practices. Further, the research suggests the need for continued work and adoption of the scheduler into production environments, especially within the context of the ING data processing pipeline.

Containerization has become a fundamental component of software development and the continuous integration and continuous delivery (CI/CD) pipeline. However, the energy overhead of containerization solutions such as Docker results not only in higher costs but also has environmental consequences. Currently, the energy efficiency of containers is often neglected, since developers mainly focus on small image size and time efficiency. Furthermore, there is also a lack of understanding about the impact of base images on the container during runtime. Therefore, the objective of this thesis is to assess the impact of base image selection on the energy efficiency of containerized workloads. This is done by running an empirical study, in which we conduct experiments with a diverse set of real-world workloads and base images. The results of this study show that the base image can have a significant impact on the energy efficiency of the container. However, the magnitude of this impact in practice depends on the workload. For some workloads (i.e. databases) this impact is more significant in practice than for others (i.e. gaming servers). While there is no single best or worst image across all workloads, Alpine is often the least energy-efficient option. Furthermore, the execution time of the container is only correlated to the energy consumption for CPU-intensive workloads. Therefore, besides image size and time efficiency, it is important to consider energy efficiency in the selection criteria as well when considering sustainable practices in software engineering.