Baransel, Berrak AlaraPeker, AlperBalkıs, Hilmi ÖmerArı, İsmail2023-05-252023-05-252021978-303071710-01865-0929http://hdl.handle.net/10679/8338https://doi.org/10.1007/978-3-030-71711-7_24Providing end-users with high quality e-commerce, online communication, education services requires careful performance monitoring, tuning and prediction under heavy traffic loads. To address this issue, we propose and evaluate a novel methodology using Docker containers for load testing. Our experience over several benchmarks, local machines vs. Cloud, and web servers suggest that load testing as a service requires a multi-dimensional optimization over slave counts, network latencies, bandwidth, and traffic patterns and there are opportunities for learning these parameters that can later be modelled into a smart load testing algorithm, with machine learning at the driver seat. Beyond the ease and speed of deployment, containers and cloud also provide a low cost alternative to load testing; we completed our cloud experiments by spending only $10. The only disadvantage of public clouds can be their centralized nature and distance to real customer bases.engrestrictedAccessconferenceObject138229230210.1007/978-3-030-71711-7_24CloudContainerDjangoDockerJmeterKubernetesLoad testing2-s2.0-85103500692