Cloud Gaming: A QoE Study of Fast-paced Single-player and Multiplayer Gaming
Sebastian Flinck Lindstrom, Markus Wetterberg, and Niklas Carlsson
Paper:
Sebastian Flinck Lindstrom, Markus Wetterberg, and Niklas Carlsson,
"Cloud Gaming: A QoE Study of Fast-paced Single-player and Multiplayer Gaming",
Proc. IEEE/ACM International Conference on Utility and Cloud Computing (IEEE/ACM UCC),
Dec. 2020.
(pdf)
Abstract:
Cloud computing offers an attractive solution for modern computer games. By moving the increasingly
demanding graphical calculations (e.g., generation of real-time video streams) to the cloud, consumers
can play games using small, cheap devices.
While cloud gaming
has many advantages and is increasingly deployed, not much work has been done to understand the
underlying factors impacting players' user experience when moving the processing to the cloud.
In this paper, we study the impact of the quality of service (QoS) factors most affecting the
players' quality of experience (QoE) and in-game performance.
In particular, these relationships are studied from multiple perspectives using complementing
analysis methods applied on the data collected via instrumented user tests.
During the tests, we manipulated the players' network conditions and collected low-level QoS
metrics and in-game performance, and after each game, the users answered questions capturing
their QoE. New insights are provided using different correlation/auto-correlation/cross-correlation
statistics, regression models, and a thorough breakdown of the QoS metric most strongly correlated
with the users' QoE. We find that the frame age is the most important QoS metric for predicting
in-game performance and QoE, and that spikes in the frame age caused by large frame transfers can
have extended negative impact as they can cause processing backlogs. The study emphasizes the need
to carefully consider and optimize the parts making up the frame age, including dependencies between
the processing steps. By lowering the frame age, more enjoyable gaming experiences can be provided.
Code and datasets
To help build upon our work, below, we make available code and example datasets.
-
Code:
The code used in our paper is made available
here (778 MB).
-
Datasets:
The primary dataset analyzed in our paper is made available
here
Additional datasets may be added later.
Note: If you use our datafiles or code in your research,
please include a reference to our UCC 2020 paper
(pdf) in your work.