Transforming Performance Engineering with Generative AI

Abstract

Generative AI is poised to revolutionize performance engineering by auto-mating key tasks, improving prediction accuracy, and enabling real-time sys-tem adaptation. This paper explores the transformative potential of integrat-ing Generative AI into performance engineering workflows, demonstrating significant improvements in test automation, system optimization, and re-al-time monitoring, while also highlighting critical challenges related to bias, security, and explainability. We examine how techniques like Large Language Models (LLMs), Reinforcement Learning (RL), and Neural Architecture Search (NAS) can address the challenges of modern application performance. Through illustrative examples, we demonstrate the benefits of AI-driven test generation, system optimization, and monitoring. We also address critical considerations such as model explainability, data privacy, and the essential role of human oversight. Finally, we outline future research directions and the long-term implications for the performance engineering landscape.