{"43ef156482932a2c6bf19b98003138f5rss":{"DOI":"","ISBN":"978-3-032-04403-7","ISSN":"","URL":"","abstract":"As software architecture grows in complexity, understanding the implications of design decisions becomes increasingly challenging. Large Language Models (LLMs) offer new opportunities for enhancing explainability during architecture modeling and evaluation by generating natural language explanations that support comprehension, learning, and decision-making. This potential is particularly valuable in domains with increased technical complexity---such as elasticity in cloud-based systems. In this work, we integrate and evaluate LLM-based explanations in supporting design-time evaluation of software architectures, focusing on the detection of elasticity antipatterns. Elasticity antipatterns are flawed autoscaling policy configurations that potentially lead to inefficient or unreliable system behavior. We extend an existing modeling and simulation approach with a novel feature that generates contextualized, textual explanations derived from simulation data. These explanations aim to guide architects in understanding scaling behaviors, identifying design issues, and refining their models. Our contribution includes the conceptualization of explanation types relevant to elasticity modeling, the design of prompt templates to elicit effective responses from LLMs, and an evaluation of the generated explanations' usefulness and quality. Results indicate that LLM-assisted feedback enhances the interpretability of elasticity models and supports the early identification of antipatterns, albeit with some limitations in precision and conciseness with only a slight agreement between expert evaluations (\\$\\$\\backslashkappa \\$\\$$\\kappa$ = 0.202). The explanation quality across types of explanations differs. Even though most explanations contain factual information, a large portion was deemed as imprecise especially in explaining problem and solution, the policy and target and service level objectives.","annote":"","author":[{"family":"Klinaku","given":"Floriment"},{"family":"Lammert","given":"Jonas"},{"family":"Becker","given":"Steffen"}],"citation-label":"10.1007/978-3-032-04403-7_14","collection-editor":[{"family":"Bianculli","given":"Domenico"},{"family":"Sartaj","given":"Hassan"},{"family":"Andrikopoulos","given":"Vasilios"},{"family":"Pautasso","given":"Cesare"},{"family":"Mikkonen","given":"Tommi"},{"family":"Perez","given":"Jennifer"},{"family":"Bures","given":"Tomás"},{"family":"De Sanctis","given":"Martina"},{"family":"Muccini","given":"Henry"},{"family":"Navarro","given":"Elena"},{"family":"Soliman","given":"Mohamed"},{"family":"Zdun","given":"Uwe"}],"collection-title":"","container-author":[{"family":"Bianculli","given":"Domenico"},{"family":"Sartaj","given":"Hassan"},{"family":"Andrikopoulos","given":"Vasilios"},{"family":"Pautasso","given":"Cesare"},{"family":"Mikkonen","given":"Tommi"},{"family":"Perez","given":"Jennifer"},{"family":"Bures","given":"Tomás"},{"family":"De Sanctis","given":"Martina"},{"family":"Muccini","given":"Henry"},{"family":"Navarro","given":"Elena"},{"family":"Soliman","given":"Mohamed"},{"family":"Zdun","given":"Uwe"}],"container-title":"Software Architecture. ECSA 2025 Tracks and Workshops","documents":[],"edition":"","editor":[{"family":"Bianculli","given":"Domenico"},{"family":"Sartaj","given":"Hassan"},{"family":"Andrikopoulos","given":"Vasilios"},{"family":"Pautasso","given":"Cesare"},{"family":"Mikkonen","given":"Tommi"},{"family":"Perez","given":"Jennifer"},{"family":"Bures","given":"Tomás"},{"family":"De Sanctis","given":"Martina"},{"family":"Muccini","given":"Henry"},{"family":"Navarro","given":"Elena"},{"family":"Soliman","given":"Mohamed"},{"family":"Zdun","given":"Uwe"}],"event-date":{"date-parts":[["2025"]],"literal":"2025"},"event-place":"Cham","id":"43ef156482932a2c6bf19b98003138f5rss","interhash":"caff878720646430bd1f1b19e87c80bf","intrahash":"43ef156482932a2c6bf19b98003138f5","issue":"","issued":{"date-parts":[["2025"]],"literal":"2025"},"keyword":"rss myown models explainability language large performance threshold-based","misc":{"isbn":"978-3-032-04403-7"},"note":"","number":"","number-of-pages":"13","page":"141--154","page-first":"141","publisher":"Springer Nature Switzerland","publisher-place":"Cham","status":"","title":"LLM-Based Explainability at Design Time: Detecting Elasticity Antipatterns in Software Architectures","type":"paper-conference","username":"rss","version":"","volume":""},"0d591f4693858eb24190f2cab15ddf25rss":{"DOI":"10.1145/3491204.3527482","ISBN":"9781450391597","ISSN":"","URL":"https://doi.org/10.1145/3491204.3527482","abstract":"Data-intensive container-based cloud applications have become popular with the increased use cases in the Internet of Things domain. Challenges arise when engineering such applications to meet quality requirements, both classical ones like performance and emerging ones like resilience. There is a lack of reference use cases, applications, and experiences when prototyping such applications that could benefit the research community. Moreover, it is hard to generate realistic and reliable workloads that exercise the resources according to a specification. Hence, designing reference applications that would exhibit similar performance behavior in such environments is hard. In this paper, we present a work in progress towards a reference use case and application for data-intensive containerized cloud applications having an industrial motivation. Moreover, to generate reliable CPU workloads we make use of ProtoCom, a well-known library for the generation of resource demands, and report the performance under various quality requirements in a Kubernetes cluster of moderate size. Finally, we present the scalability of the current solution assuming a particular autoscaling policy. Results of the calibration show high variability of the ProtoCom library when executed in a cloud environment. We observe a moderate association between the occupancy of node and the relative variability of execution time.","annote":"","author":[{"family":"Klinaku","given":"Floriment"},{"family":"Rapp","given":"Martina"},{"family":"Henss","given":"Jörg"},{"family":"Rhode","given":"Stephan"}],"citation-label":"Klinaku2022","collection-editor":[],"collection-title":"ICPE '22","container-author":[],"container-title":"Companion of the 2022 ACM/SPEC International Conference on Performance Engineering","documents":[],"edition":"","editor":[],"event-date":{"date-parts":[["2022","7","14"]],"literal":"2022"},"event-place":"Bejing, China","id":"0d591f4693858eb24190f2cab15ddf25rss","interhash":"9cdf2b74ba4ff7d4431df5f835d0d356","intrahash":"0d591f4693858eb24190f2cab15ddf25","issue":"","issued":{"date-parts":[["2022","7","14"]],"literal":"2022"},"keyword":"myown from:klinaku cloud-computing performance threshold-based","misc":{"isbn":"9781450391597","location":"Bejing, China","doi":"10.1145/3491204.3527482"},"note":"","number":"","number-of-pages":"7","page":"53–60","page-first":"53","publisher":"Association for Computing Machinery","publisher-place":"Bejing, China","status":"","title":"Beauty and the Beast: A Case Study on Performance Prototyping of Data-Intensive Containerized Cloud Applications","type":"paper-conference","username":"rss","version":"","volume":""}}