Game action based power management for multiplayer online game

Abstract

Current mobile devices embrace a wide range of functionalities including high speed network support, hardware accelerated 3D graphics, and multimedia capabilities. These capabilities have boosted the interest for enabling multiplayer online games (MOG) support on such devices. However, the lack of similar growth in battery technology limits the usability of these devices for MOGs. In this paper, we present energy conservation techniques for highly interactive MOGs. These are games, such as firstperson shooters, where crisp user interaction is paramount to the overall game experience. Hence, conserving energy while preserving crisp user interaction becomes a critical consideration in this domain. We first present three obvious power management approaches and highlight their limitations. We then discuss two applicationassisted approaches for power management that manage to save power while preserving the required user experience. Our results demonstrate that these applicationassisted approaches are very promising. Categories and Subject Descriptors General TermsAlgorithms, Management, Measurement, Performance, Design, Reliability, Experimentation, Human Factors. KeywordsMobile games, wireless networks, power management, statistical prediction. INTRODUCTIONThe number of personal portable devices sold each year is increasing rapidly with mobile phone sales outpacing personal computer sales at the rate of 5 to 1 7. Current mobile phones are not just devices for voice communication, but are mini computers that have high speed wired and wireless networks, and sufficient processing and memory capabilities to run enterprise, multimedia, and gaming applications. However, the user experience obtained from running multimedia streaming and highly interactive applications on mobile phones is usually much lower when compared to running those same applications on a desktop computer. Beside several user interface factors, there are several technical factors that hinder the user experience for multimedia streaming and interactive applications on mobile phones. One of the key challenges is battery lifetime. For example, a mobile phone's battery may only be able to support 4 to 6 hours of actual voice communications and about 2 to 4 hours of streaming video. This perilous energy situation becomes even worse when the 802.11b or g network interface of the mobile phone is activated. 802.11b/g has higher energy usage compared to GSM based networks as the Carrier Sense function of 802.11b/g's CSMA/CA protocol consumes energy continuously. As various studies have shown 77, the wireless interface and CPU are the major power consuming components in mobile devices the wireless interface alone can consume up to 50% of the total system power.In this paper we first present three obvious power management approaches and demonstrate their limitations. We then present two approaches based on predicting a game player's behavior. In these approaches, we first use a simple linear prediction algorithm to decide the player's activity level and ...

Time Series Characterization of Gaming Workload for Runtime Power Management

“…On the other hand, the graphics processing pipeline and the architecture of graphics processors were also exploited in [38] for power management of games. In a different direction, techniques that rely on user-interactions [4] or require the user to manually switch to a higher clock frequency whenever a drop in game quality is observed [24], [28] have also been proposed.…”

“…Quake II was chosen for two reasons: i) We could compare our results to other related work that studied Quake II [4], [17], [20]. ii) Quake II uses software-based rendering and its engine forms the core of a variety of other games as mentioned above.…”

ARIVU: Power-aware middleware for multiplayer mobile games

“…These actions closely correlate with the amount of game-state updates required to maintain an acceptable end-user game experience. Our experiments and analysis about the effect of the game actions on the game-state updates with Quake-mobile is described in our previous work [1].…”

“…PAL is defined as the number of keyboard or mouse (or other UI devices) activities per second. Higher PAL rates indicate more critical actions.The effect of PAL on game state is described in more detail in our previous work [1].…”

“…Through set of experiments with players of different skill levels we found that past 5 actions are sufficient to achieve more than 90% accuracy. We currently use initial weights of 1 2 , 1 4 , 1 8 , 1 16 and 1 16 for w 0 to w 4 respectively so that ∑ j w j = 1. The weights can be dynamically adjusted using the prediction error to improve the prediction accuracy.…”