Under CopyrightSeeliger, RobertRobertSeeligerMüller, ChristophChristophMüllerArbanowski, StefanStefanArbanowski2022-12-192022-12-192022https://publica.fraunhofer.de/handle/publica/430158https://doi.org/10.24406/publica-66110.1109/IoTaIS56727.2022.997591910.24406/publica-661With the growing usage of high quality HD and ultra HD video content, adaptive bitrate streaming and constantly increasing demand for bitrates and distribution bandwidth, energy consumption and related costs grow exponentially in parallel. As such, it is vital to reduce the overall energy consumption of online video streaming. In this paper we aim to investigate, which parameters influence energy consumption for video streaming, on the client (device) side, as well as during encoding. To conduct this systematic investigation, we have set up a reproducible measurement environment that closely resembles real-world conditions, with different client devices, and video encoding workflows, each connected to energy measurement devices. In an advanced step, we additionally examine the effect of content aware encoding methods on power consumption, using an AI-based perscene encoding solution. Finally, we discuss and evaluate the measurements and offer recommendations to reduce overall CO2 emissions for video streaming.engreen communicationsenergy-efficiencyvideo encodingmachine learningstreaming workflow optimizationadaptive bitrate streamingDDC::300 Sozialwissenschaften::330 Wirtschaft::338 ProduktionDDC::000 Informatik, Informationswissenschaft, allgemeine Werke::000 Informatik, Wissen, Systeme::006 Spezielle Computerverfahrenconference paper