FUTURE WORK While the server-side measurements are still more or less work in progress, we can already give some estimates about it. As the most minimal configuration, we tested that a Raspberry Pi running a game similar to OpenArena (Quake 3) would consume only 3 watts. We were unfortunately unable to run GamingAnywhere successfully on the ARM platform. In contrast, the mini PC consumed roughly 20 watts in idle mode and an extra 5 watts with OpenArena and GamingAnywhere running on it. During the review process of this paper, we have already virtualized the server side by using docker-based containers. While measuring of the power consumption of the server side is still work in progress, we would estimate that Docker would add a maximum of two to three watts overhead in the case of the mini PC. Some recent studies about container performance show how container energy overhead – compared to non-virtualized applications – is almost negligible for a wide range of workloads. For example, Morabito [14] reports that running eight containers on a single server increases the power consumption only by one watt. Moreover, the same study shows that container based virtualization is more powerefficient when compared to hypervisor-based virtualization. Various other miscellaneous aspects would also be interesting as future work items. It would beneficial to compare the energy efficiency of GamingAnywhere with Splashtop remote desktop [10], which is apparently also suitable for games. It would be also useful to measure how close to the end-user a distributed game cloud solution should be deployed; maybe a regional datacenter could be enough instead of co-locating the game servers near the network access point.