2.4 Energy Efficient Mobile Platforms
2.4.6 Energy Efficient Applications
2.4.6.2 Adaptive Control at Application Layer
In order to optimize energy efficiency and improve user experience, applications can be designed to dynamically adapt the streaming process to network conditions, observed power levels or user-side quality measures, as shown in Fig.2.11.
Recent studies have been focused on interactive compression techniques where the sender adaptively chooses the optimal compression rate to meet energy efficiency or QoS constraints.
The solution presented in [139] adapts the compression strategy based on feedback from clients and guarantees the constrained minimum QoS. The battery-powered client device sends the server its maximum decoding capability, so the server could calculate the optimal transmission rate. It is suggested that if the decoding aptitude (P) matches the number of correctly received packets at the client side (Q), the client will achieve the best energy efficiency. If the client receives more packets than it can decode in real time, the energy spent on receiving Q-P packets is wasted. On the other hand if Q is smaller than P, the server should send more packets to improve video quality. Thus the feedback mechanism helps balance these two values and achieve energy efficiency.
Quality-Oriented Adaptation Scheme (QOAS) proposed in [140] uses feedback received from clients, which mainly includes user perceived quality and QoS param- eters such as average loss rate, to dynamically adjust the streaming rate at the server side. Simulation results show significant increase in the number of clients that can be served simultaneously and meanwhile the quality of service is maintained at high level.
The mechanism proposed in [141] selects the optimal image compression parame- ters at runtime to best balance the tradeoff between energy, latency and image quality. The methodology consists of two steps. In the first step, the average value of the image quality and latency is calculated. In the second step, the data obtained from the first step is used to generate a table with quality and latency constraints and total energy consumption spent on computing and transmitting images. The table is then used to look up the optimal parameters for the desired energy/latency/image quality.
Adaptive Source Rate Control (ASRC) [142] for video streaming applications is proposed to work with hybrid Automatic Repeat Request (ARQ). It takes advantage of high throughput and reliability achieved by ARQ and at the same time guarantees that data could get to the destination within the delay limit. ACK packets received at the data source are used to calculate packet error rate which is an indicator of channel condition. the ASRC scheme forecasts the channel effective data rate based on the
error rate before the next video frame is encoded. Finally the target number of bits for the next frame is calculated according to the channel conditions and the target delay limit so that the packets can be transmitted correctly and within the imposed delay limits.
The solution presented in [143] adjusts video streaming strategy at runtime to pro- long service time of a whole wireless communication system. The authors observed that the video quality is determined by three aspects: encoding aptitude of the server, decoding aptitude of the client, and the channel. Therefore they propose a strategy where transmission power level at the server side and decoding scheme at the client side is adjusted at each frame based on the energy level at runtime with guaranteed minimum video quality. The adjustment is made with consideration of energy level on both sides as the system life time is maximized if the server and client run out of energy at the same time.
EVAN [144] and ESTREL [145] use this approach, adapting the video quality based on device characteristics and remaining battery levels. Scalable video coding such as MPEG-4 SVC [146] enables layer-based multimedia quality adjustments. De- vices subscribe to enhancement layers only if their remaining energy levels are high. Otherwise they un-subscribe from some enhancement layers to reduce the amount of data to be received/transmitted and save energy. SAMMy [147] is a dynamic video delivery solution that adjusts content quality based on estimated signal strength and monitored packet loss rate. These parameters are utilised to make more efficient use of the wireless network resources, increase user perceived quality and save energy. DEAS [148] adaptively changes the video QoS level by monitoring the application holding on and the current residual energy. DEAS is the first adaptive streaming so- lution that considers application running environment (i.e. not only the current multi- media streaming application, but also other applications) and device features that put different energy constraints on the device. Alt et al. have proposed [149] that assess the level of movement between continuous frames. The frames with major difference
than the previous frame have to be delivered as important information is lost other- wise. However it drops frames with little difference in movement to save energy. Park et al. [150] have proposed a SNR scalable architecture that trans-coding from H.264 to SVC for energy saving. They developed a dedicated chip for trans-coding in order to release mobile CPU from the computational complexity of trans-coding in adaptive content delivery scenario.
The power-aware scheme introduced in [151] dynamically adapts the behaviour of applications according to energy levels in order to prolong battery life for mobile devices. High quality of service is achieved if the battery resource is plentiful. Energy conservation is performed at the expense of user experience when a device is running out of energy. Experiments based on four different application types: video player, speech recognizer, map viewer and web browser are performed and testing results show that lowering data fidelity yields significant energy savings.
Figure 2.12: Illustration of Partial Caching.