Philipp Leitner

Hazem Peter Samoaa, Antonio Longa, Mazen Mohamad et al.

Predicting the performance of production code prior to actually executing or benchmarking it is known to be highly challenging. In this paper, we propose a predictive model, dubbed TEP-GNN, which demonstrates that high-accuracy performance prediction is possible for the special case of predicting unit test execution times. TEP-GNN uses FA-ASTs, or flow-augmented ASTs, as a graph-based code representation approach, and predicts test execution times using a powerful graph neural network (GNN) deep learning model. We evaluate TEP-GNN using four real-life Java open source programs, based on 922 test files mined from the projects' public repositories. We find that our approach achieves a high Pearson correlation of 0.789, considerable outperforming a baseline deep learning model. However, we also find that more work is needed for trained models to generalize to unseen projects. Our work demonstrates that FA-ASTs and GNNs are a feasible approach for predicting absolute performance values, and serves as an important intermediary step towards being able to predict the performance of arbitrary code prior to execution.

Huaifeng Zhang, Fahmi Abdulqadir Ahmed, Dyako Fatih et al.

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%.

Peter Samoaa, Linus Aronsson, Antonio Longa et al.

Most machine learning and data analytics applications, including performance engineering in software systems, require a large number of annotations and labelled data, which might not be available in advance. Acquiring annotations often requires significant time, effort, and computational resources, making it challenging. We develop a unified active learning framework specializing in software performance prediction to address this task. We begin by parsing the source code to an Abstract Syntax Tree (AST) and augmenting it with data and control flow edges. Then, we convert the tree representation of the source code to a Flow Augmented-AST graph (FA-AST) representation. Based on the graph representation, we construct various graph embeddings (unsupervised and supervised) into a latent space. Given such an embedding, the framework becomes task agnostic since active learning can be performed using any regression method and query strategy suited for regression. Within this framework, we investigate the impact of using different levels of information for active and passive learning, e.g., partially available labels and unlabeled test data. Our approach aims to improve the investment in AI models for different software performance predictions (execution time) based on the structure of the source code. Our real-world experiments reveal that respectable performance can be achieved by querying labels for only a small subset of all the data.

David Schön, Faiza Amjad, Tehreem Asif et al.

Large language models (LLMs) have gained widespread popularity and have steadily improved over time, enabling software developers to use them for various code-related tasks. One common task is code refactoring, where the LLM suggests changes for the developer to apply to their code to improve quality attributes such as readability or maintainability. While current research focuses on evaluating LLM-generated refactoring suggestions, there is a limited understanding of how developers apply these suggestions in practice. To explore this, we analyze 169 GitHub commits where developers refactor their code based on a ChatGPT conversation linked in the commit message. We found that developers mostly accept and use the suggestions without modifications. When changes are made, they are mostly major and fall into five different patterns that depend on the refactoring activity, the developer's prompt, and the validity of the response from ChatGPT.

Lirong Yi, Gregory Gay, Philipp Leitner

Large Language Models (LLMs) can generate code, but can they generate fast code? In this paper, we study this question using a dataset of 65 real-world tasks mined from open-source Java programs. We specifically select tasks where developers achieved significant speedups, and employ an automated pipeline to generate patches for these issues using two leading LLMs under four prompt variations. By rigorously benchmarking the results against the baseline and human-authored solutions, we demonstrate that LLM-generated code indeed improves performance over the baseline in most cases. However, patches proposed by human developers outperform LLM fixes by a statistically significant margin, indicating that LLMs often fall short of finding truly optimal solutions. We further find that LLM solutions are semantically identical or similar to the developer optimization idea in approximately two-thirds of cases, whereas they propose a more original idea in the remaining one-third. However, these original ideas only occasionally yield substantial performance gains.

Linda Erlenhov, Francisco Gomes de Oliveira Neto, Philipp Leitner

Software engineering bots - automated tools that handle tedious tasks - are increasingly used by industrial and open source projects to improve developer productivity. Current research in this area is held back by a lack of consensus of what software engineering bots (DevBots) actually are, what characteristics distinguish them from other tools, and what benefits and challenges are associated with DevBot usage. In this paper we report on a mixed-method empirical study of DevBot usage in industrial practice. We report on findings from interviewing 21 and surveying a total of 111 developers. We identify three different personas among DevBot users (focusing on autonomy, chat interfaces, and "smartness"), each with different definitions of what a DevBot is, why developers use them, and what they struggle with. We conclude that future DevBot research should situate their work within our framework, to clearly identify what type of bot the work targets, and what advantages practitioners can expect. Further, we find that there currently is a lack of general purpose "smart" bots that go beyond simple automation tools or chat interfaces. This is problematic, as we have seen that such bots, if available, can have a transformative effect on the projects that use them.

Ranim Khojah, Mazen Mohamad, Philipp Leitner et al.

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 (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.

Ranim Khojah, Francisco Gomes de Oliveira Neto, Mazen Mohamad et al.

Large Language Models (LLMs) are increasingly used by software engineers for code generation. However, limitations of LLMs such as irrelevant or incorrect code have highlighted the need for prompt programming (or prompt engineering) where engineers apply specific prompt techniques (e.g., chain-of-thought or input-output examples) to improve the generated code. While some prompt techniques have been studied, the impact of different techniques -- and their interactions -- on code generation is still not fully understood. In this study, we introduce CodePromptEval, a dataset of 7072 prompts designed to evaluate five prompt techniques (few-shot, persona, chain-of-thought, function signature, list of packages) and their effect on the correctness, similarity, and quality of complete functions generated by three LLMs (GPT-4o, Llama3, and Mistral). Our findings show that while certain prompt techniques significantly influence the generated code, combining multiple techniques does not necessarily improve the outcome. Additionally, we observed a trade-off between correctness and quality when using prompt techniques. Our dataset and replication package enable future research on improving LLM-generated code and evaluating new prompt techniques.

Ranim Khojah, Francisco Gomes de Oliveira Neto, Philipp Leitner

Software developers use natural language to interact not only with other humans, but increasingly also with chatbots. These interactions have different properties and flow differently based on what goal the developer wants to achieve and who they interact with. In this paper, we aim to understand the dynamics of conversations that occur during modern software development after the integration of AI and chatbots, enabling a deeper recognition of the advantages and disadvantages of including chatbot interactions in addition to human conversations in collaborative work. We compile existing conversation attributes with humans and NLU-based chatbots and adapt them to the context of software development. Then, we extend the comparison to include LLM-powered chatbots based on an observational study. We present similarities and differences between human-to-human and human-to-bot conversations, also distinguishing between NLU- and LLM-based chatbots. Furthermore, we discuss how understanding the differences among the conversation styles guides the developer on how to shape their expectations from a conversation and consequently support the communication within a software team. We conclude that the recent conversation styles that we observe with LLM-chatbots can not replace conversations with humans due to certain attributes regarding social aspects despite their ability to support productivity and decrease the developers' mental load.

Ranim Khojah, Mazen Mohamad, Linda Erlenhov et al.

The risks associated with adopting large language model (LLM) chatbots in software organizations highlight the need for clear policies. We examine how 11 companies create these policies and the factors that influence them, aiming to help managers safely integrate chatbots into development workflows.

Peter Samoaa, Mehrdad Farahani, Antonio Longa et al.

The landscape of deep learning has vastly expanded the frontiers of source code analysis, particularly through the utilization of structural representations such as Abstract Syntax Trees (ASTs). While these methodologies have demonstrated effectiveness in classification tasks, their efficacy in regression applications, such as execution time prediction from source code, remains underexplored. This paper endeavours to decode the behaviour of tree-based neural network models in the context of such regression challenges. We extend the application of established models--tree-based Convolutional Neural Networks (CNNs), Code2Vec, and Transformer-based methods--to predict the execution time of source code by parsing it to an AST. Our comparative analysis reveals that while these models are benchmarks in code representation, they exhibit limitations when tasked with regression. To address these deficiencies, we propose a novel dual-transformer approach that operates on both source code tokens and AST representations, employing cross-attention mechanisms to enhance interpretability between the two domains. Furthermore, we explore the adaptation of Graph Neural Networks (GNNs) to this tree-based problem, theorizing the inherent compatibility due to the graphical nature of ASTs. Empirical evaluations on real-world datasets showcase that our dual-transformer model outperforms all other tree-based neural networks and the GNN-based models. Moreover, our proposed dual transformer demonstrates remarkable adaptability and robust performance across diverse datasets.

Hamdy Michael Ayas, Philipp Leitner, Regina Hebig

Background: Microservices migrations are challenging and expensive projects with many decisions that need to be made in a multitude of dimensions. Existing research tends to focus on technical issues and decisions (e.g., how to split services). Equally important organizational or business issues and their relations with technical aspects often remain out of scope or on a high level of abstraction. Aims: In this study, we aim to holistically chart the decision-making that happens on all dimensions of a migration project towards microservices (including, but not limited to, the technical dimension). Method: We investigate 16 different migration cases in a grounded theory interview study, with 19 participants that recently migrated towards microservices. This study strongly focuses on the human aspects of a migration, through stakeholders and their decisions. Results: We identify 3 decision-making processes consisting of 22decision-points and their alternative options. The decision-points are related to creating stakeholder engagement and assessing feasibility, technical implementation, and organizational restructuring. Conclusions: Our study provides an initial theory of decision-making in migrations to microservices. It also outfits practitioners with a roadmap of which decisions they should be prepared to make and at which point in the migration.

Rostyslav Zabolotnyi, Philipp Leitner, Waldemar Hummer et al.

The Infrastructure-as-a-Service (IaaS) model of cloud computing is a promising approach towards building elastically scaling systems. Unfortunately, building such applications today is a complex, repetitive and error-prone endeavor, as IaaS does not provide any abstraction on top of naked virtual machines. Hence, all functionality related to elasticity needs to be implemented anew for each application. In this paper, we present JCloudScale, a Java-based middleware that supports building elastic applications on top of a public or private IaaS cloud. JCloudScale allows to easily bring applications to the cloud, with minimal changes to the application code. We discuss the general architecture of the middleware as well as its technical features, and evaluate our system with regard to both, user acceptance (based on a user study) and performance overhead. Our results indicate that JCloudScale indeed allowed many participants to build IaaS applications more efficiently, comparable to the convenience features provided by industrial Platform-as-a-Service (PaaS) solutions. However, unlike PaaS, using JCloudScale does not lead to a loss of control and vendor lock-in for the developer.

Jürgen Cito, Philipp Leitner, Thomas Fritz et al.

Cloud computing is gaining more and more traction as a deployment and provisioning model for software. While a large body of research already covers how to optimally operate a cloud system, we still lack insights into how professional software engineers actually use clouds, and how the cloud impacts development practices. This paper reports on the first systematic study on how software developers build applications in the cloud. We conducted a mixed-method study, consisting of qualitative interviews of 25 professional developers and a quantitative survey with 294 responses. Our results show that adopting the cloud has a profound impact throughout the software development process, as well as on how developers utilize tools and data in their daily work. Among other things, we found that (1) developers need better means to anticipate runtime problems and rigorously define metrics for improved fault localization and (2) the cloud offers an abundance of operational data, however, developers still often rely on their experience and intuition rather than utilizing metrics. From our findings, we extracted a set of guidelines for cloud development and identified challenges for researchers and tool vendors.

Joel Scheuner, Philipp Leitner, Jurgen Cito et al.

To optimally deploy their applications, users of Infrastructure-as-a-Service clouds are required to evaluate the costs and performance of different combinations of cloud configurations to find out which combination provides the best service level for their specific application. Unfortunately, benchmarking cloud services is cumbersome and error-prone. In this paper, we propose an architecture and concrete implementation of a cloud benchmarking Web service, which fosters the definition of reusable and representative benchmarks. In distinction to existing work, our system is based on the notion of Infrastructure-as-Code, which is a state of the art concept to define IT infrastructure in a reproducible, well-defined, and testable way. We demonstrate our system based on an illustrative case study, in which we measure and compare the disk IO speeds of different instance and storage types in Amazon EC2.