Improving Energy Efficiency of Virtual Machines with Timer Tick Variations

Amount of media content viewed by users is impacting the power consumption and thereby reducing battery life of mobile platforms. Increased use of virtual machines complicates diagnosis of underlying causes of this increased power consumption, as multiple operating systems are interacting with each other. We identify processor state residencies using software tools and instrument the target device to measure platform component power. The interrupt and context switch rate are studied to relate their influence on power consumption. Later, we propose a timer tick varying technique which reduces the interaction between virtual machines at idle state. This is verified to reduce power consumption by 3 W. Context switch rate and interrupt rate are reduced by 23% and 34% respectively by this optimization.