Gortázar, FranciscoGallego, MicaelMaes-Bermejo, MichelChicano-Capelo, IvánSantos, Carlos2023-09-212023-09-212022Francisco Gortázar, Micael Gallego, Michel Maes-Bermejo, Iván Chicano-Capelo, Carlos Santos, Cost-effective load testing of WebRTC applications, Journal of Systems and Software, Volume 193, 2022, 111439, ISSN 0164-1212, https://doi.org/10.1016/j.jss.2022.1114391873-1228https://hdl.handle.net/10115/24453A repository with the OpenVidu Loadtest Tool repository is available in GitHub.13 Additionally, a reproduction package (Bellas et al., 2021) is available in Zenodo with raw results and a copy of the OpenVidu Loadtest Tool repository. The authors would like to thank the reviewers for their insightful comments that helped improve the paper. This work has been supported by the Government of Spain through project “BugBirth” (RTI2018-101963-B-100), by the Regional Government of Madrid (Spain) (CM) through project EDGEDATA-CM (P2018/TCS-4499) cofunded by FSE & FEDER, and by the European Commission under the H2020 project “MICADO” (GA-822717).Background: Video conference applications and systems implementing the WebRTC W3C standard are becoming more popular and demanded year after year, and load testing them is of paramount importance to ensure they can cope with demand. However, this is an expensive activity, usually involving browsers to emulate users. Goal : to propose browser-less alternative strategies for load testing WebRTC services, and to study performance and costs of those strategies when compared with traditional ones. Method: (a) Exploring the limits of existing and novel strategies for load testing WebRTC services from a single machine. (b) Comparing the common strategy of using browsers with the best of our proposed strategies in terms of cost in a load testing scenario. Results: We observed that, using identical machines, our proposed strategies are able to emulate more users than traditional strategies. We also found a huge saving in expenditure for load testing, as our strategy suppose a saving of 96% with respect to usual browser-based strategies. We also found there are almost no differences between the traditional strategies considered. Conclusion: We provide details on scalability of different load testing strategies in terms of users emulated, as well as CPU and memory used. We could reduce the expenditure of load tests of WebRTC applications.engAtribución 4.0 Internacionalhttp://creativecommons.org/licenses/by/4.0/TestingLoad testingWebRTCinfo:eu-repo/semantics/article10.1016/j.jss.2022.111439info:eu-repo/semantics/openAccess