Outline of our research direction
Investigate impact on gaming quality
- specifically analyze the trade-off between energy savings and gaming quality.
- find paper to model gaming quality.
Improve algorithm from linear model to exponential curve.
- compare performance between existing solutions, linear, and exponential model.
Expand simulation/implementation to different games.
- find reference to prove DR is used in practice.
- Show that our solution is game independent
Expand analysis to include a global view.
- mathematical model
-Performance metric in terms of bandwidth savings, can be calculated from a global view perspective.
Solve receiving mode problems
- Use PSM for AP mode, and rotating host algorithm for ad-hoc.
- Please see Yi's thesis work.
Additional Ideas
- slow down the game to XX fps to allow the game to sleep longer. Without affecting gaming quality.
- see Claypool's papers regarding frame rates and gaming quality.
Energy Savings Overhead
- Investigate energy constraints for nodes with and without our energy savings algorithm.
- fairness in game play/quality.
- have nodes without energy constraints do more work
Other Notes
- Continue to be vague regarding the specific wireless standards. ie. we don't target WiFi or 3G. Instead, we let the reader decides.
- Discover if there is any periodicity of the data and use this to help determine sleep cycles
- Provide strong evidence that our solution helps
To Do's
- Address the items above ^
- Simulate energy savings using Carson's GLS simulator, please talk to Cameron.
- We have also modified the GLS code to replay latency traces. Please read our IRS MM'10 paper.
- Implement the algorithm into BZFlag.
- Play a LAN game setup (2 or 3 players) to test the algorithm in energy savings mode
- Play over the Internet against other players to test algorithm in a mixed environment (with nodes without energy savings)
- Play over a wireless connection with PSM mode enabled (must use an AP)
- Play over an Ad-Hoc network setup.
- If possible, find an Android open source game and implement the algorithm.
- Publish many papers :)
Other ideas
- If DR is not feasible, most of the items above can be applied to other types of games that do not use DR. Such as:
- Turn-based games (very long silent periods)
- Real-time strategy (RTS) games (medium silent periods)
- Simulation games (medium silent periods, potentially large packet sizes)
- MMORPG games (World of Warcraft client will soon be available on the Apple iPad)
- Virtual Reality, such as Second Life.
Addressing NOSSDAV'10 Reviewers Comments
- Should we use a cummulative moving average or an exponential smoothing one (similar to TCP's RTT estimation)? Moving averages assign equal weights to the (i.e. 1/N).
- We need to quantify the effect/cost of losing updates (while in sleep mode) on game performance (e.g. precision while shooting). In this direction, we can follow an approach similar to that of Claypool's MMSys'10 paper. This effect can also be added to the PMF figure in the paper.
- Elaborate on the reason behind the spike at 0.1 average error rate in the Fig. 5 bar chart.
- Base our energy consumption model on a more recent WNIC.
References
Please check these references.
- Game Action Based Power Management for Multiplayer Online Game, Anand et al.
- Experiences from Implementing a Mobile Multiplayer Real-Time Game for Wireless Networks with High Latency, Wang et al.
- Games are Up for DVFS, Gu et al.
- M. Claypool and K. Claypool, "Latency Can Kill: Precision and Deadline in Online Games", MMSys'10, 2010
- Fairness in Dead-Reckoning based Distributed Multiplayer Games, Aggarwal et al.
- A Distributed Power Management Policy for Wireless AdHoc Networks, Chiasserini et al.
- Energy Consumption in Mobile Phones: A Measurement Study..., Balasubramanian et al. (this is a good starting point to build the testbed)
- Quake3 port to Android
- PowerTutor an energy profiler for Android phones developed by University of Michigan (source code not provided)
