Power-Aware Cloud Computing Infrastructure for Latency-Sensitive Internet-of-Things Services

With the developing of Cloud Computing technology more and more latency-sensitive critical operations in Internet of Things are taken over by Cloud infrastructure. Meeting the constraints of real-time and power-aware is important for the deployment of Cloud. While various strategies have been developed to reduce power consumption of hardware components by transitioning them to lower power states or hibernating them when workload is relative low, they fail to work for latency-sensitive applications in virtualization environment. This paper presents a novel paravirtualized power-aware hypervisor as Cloud Computing infrastructure fits for latency-sensitive applications. This hypervisor introduces latency measurement process in guest OS side and CPU Power Modulator in host OS side. It enables the host OS modulate the CPU frequency level to save power with latency guarantee in guest OS side. The benchmark results show that this approach solidly improves Cloud Computing infrastructure power efficiency for latency-sensitive services.