Internet Computing and Emerging Technologies lab (ICET-lab)

Modern software development increasingly involves the use of large language models (LLMs) to generate code. Despite their rapid advancement, LLMs remain prone to errors and hallucinations, emphasizing the importance of careful code inspection. However, in practice, developers’ trust in LLM-generated code and their willingness to review it thoroughly may differ from these recommendations. In this study, we conduct a Wizard-of-Oz experiment to examine how software engineers behave when code is explicitly labeled as LLM-generated during a code review task. We collect both behavioral data and participant feedback through eye-tracking and exit interviews. Combining Bayesian data analysis with qualitative analysis, we found that while the thoroughness of code review did not change for participants, they spent more time fixating on LLM-labelled code, indicating that the label itself influences attention. Practitioners also adapted their review strategy for LLM-labelled code by assessing the code based on specific criteria (e.g., logical correctness), or using the prompt to guide their review. Our study reveals a gap between reviewers’ intentions and actual reviewing behaviour, highlighting the need for software companies to revisit their AI policies to better support developers in reviewing LLM-generated code.

Making serverless computing widely applicable requires detailed performance understanding. Although contemporary benchmarking approaches exist, they report only coarse results, do not apply distributed tracing, do not consider asynchronous applications, and provide limited capabilities for (root cause) analysis. Addressing this gap, we design and implement ServiBench, a serverless benchmarking suite. ServiBench (i) leverages synchronous and asynchronous serverless applications representative of production usage, (ii) extrapolates cloud-provider data to generate realistic workloads, (iii) conducts comprehensive, end-to-end experiments to capture application-level performance, (iv) analyzes results using a novel approach based on (distributed) serverless tracing, and (v) supports comprehensively serverless performance analysis. With ServiBench, we conduct comprehensive experiments on AWS, covering five common performance factors: median latency, cold starts, tail latency, scalability, and dynamic workloads. We find that the median end-to-end latency of serverless applications is often dominated not by function computation but by external service calls, orchestration, or trigger-based coordination. We release collected experimental data under FAIR principles and ServiBench as a tested, extensible open-source tool.

Large Language Models (LLMs) are frequently discussed in academia and the general public as support tools for virtually any use case that relies on the production of text, including software engineering. Currently, there is much debate, but little empirical evidence, regarding the practical usefulness of LLM-based tools such as ChatGPT for engineers in industry. We conduct an observational study of 24 professional software engineers who have been using ChatGPT over a period of one week in their jobs, and qualitatively analyse their dialogues with the chatbot as well as their overall experience (as captured by an exit survey). We find that rather than expecting ChatGPT to generate ready-to-use software artifacts (e.g., code), practitioners more often use ChatGPT to receive guidance on how to solve their tasks or learn about a topic in more abstract terms. We also propose a theoretical framework for how the (i) purpose of the interaction, (ii) internal factors (e.g., the user’s personality), and (iii) external factors (e.g., company policy) together shape the experience (in terms of perceived usefulness and trust). We envision that our framework can be used by future research to further the academic discussion on LLM usage by software engineering practitioners, and to serve as a reference point for the design of future empirical LLM research in this domain.

Today’s software is bloated with both code and features that are not used by most users. This bloat is prevalent across the entire software stack, from operating systems and applications to containers. Containers are lightweight virtualization technologies used to package code and dependencies, providing portable, reproducible and isolated environments. For their ease of use, data scientists often utilize machine learning containers to simplify their workflow. However, this convenience comes at a cost: containers are often bloated with unnecessary code and dependencies, resulting in very large sizes. In this paper, we analyze and quantify bloat in machine learning containers. We develop MMLB, a framework for analyzing bloat in software systems, focusing on machine learning containers. MMLB measures the amount of bloat at both the container and package levels, quantifying the sources of bloat. In addition, MMLB integrates with vulnerability analysis tools and performs package dependency analysis to evaluate the impact of bloat on container vulnerabilities. Through experimentation with 15 machine learning containers from TensorFlow, PyTorch, and Nvidia, we show that bloat accounts for up to 80% of machine learning container sizes, increasing container provisioning times by up to 370% and exacerbating vulnerabilities by up to 99%.

Performance is a crucial non-functional requirement of many software systems. Despite the widespread use of performance testing, developers still struggle to construct and evaluate the quality of performance tests. To address these two major challenges, we implement a framework, dubbed ju2jmh, to automatically generate performance microbenchmarks from JUnit tests and use mutation testing to study the quality of generated microbenchmarks. Specifically, we compare our ju2jmh generated benchmarks to manually written JMH benchmarks and to automatically generated JMH benchmarks using the AutoJMH framework, as well as directly measuring system performance with JUnit tests. For this purpose, we have conducted a study on three subjects (Rxjava, Eclipse-collections, and Zipkin) with sim∼454K source lines of code (SLOC), 2,417 JMH benchmarks (including manually written and generated AutoJMH benchmarks) and 35,084 JUnit tests. Our results show that the ju2jmh generated JMH benchmarks consistently outperform using the execution time and throughput of JUnit tests as a proxy of performance and JMH benchmarks automatically generated using the AutoJMH framework while being comparable to JMH benchmarks manually written by developers in terms of tests’ stability and ability to detect performance bugs. Nevertheless, ju2jmh benchmarks are able to cover more of the software applications than manually written JMH benchmarks during the microbenchmark execution. Furthermore, ju2jmh benchmarks are generated automatically, while manually written JMH benchmarks require many hours of hard work and attention; therefore our study can reduce developers’ effort to construct microbenchmarks. In addition, we identify three factors (too low test workload, unstable tests and limited mutant coverage) that affect a benchmark’s ability to detect performance bugs. To the best of our knowledge, this is the first study aimed at assisting developers in fully automated microbenchmark creation and assessing microbenchmark quality for performance testing.