Minimizing energy dissipation in content distribution networks using dynamic power management

The growing end-user demand for video services with superior quality on laptops, tablets, and smartphones spurs the deployment of telco content distribution networks (CDNs). Such CDNs provide scalable and bandwidth-efficient video delivery thanks to disk-packed cache servers deployed in the telco's data centers near the clients. However, a sustainable growth of these CDNs may be hindered by their lack of energy proportionality. In this paper we propose to apply dynamic power management (DPM) to the CDN's cache servers and their disks to increase the CDN's energy efficiency. DPM adapts the power states of caches and disks, which determine the overall CDN power state, to the time-varying workload. We present an offline heuristic algorithm that searches per time interval for the CDN power state that minimizes the energy consumption while avoiding an unacceptable deterioration of the CDN's performance. The algorithm guarantees sufficient delivery capacity while limiting the aggregate data rate from the origin to the caches. We evaluate the algorithm using a CDN energy simulator driven by HTTP adaptive-streaming workload traces recorded by an operational CDN delivering IPTV to mobile devices. We avoid an exhaustive search through the CDN power state space because the number of different CDN power states is exponential in the number of caches. Instead we use a greedy algorithm that exhibits only polynomial time complexity while approximating the optimal CDN power state per time interval. In addition, we show that we don't need to run a simulation for every CDN power state the search algorithm explores. To shorten the total time required for simulation even further, we run simulations in parallel. Even for a minimally-provisioned CDN, we observe a reduction of the energy dissipation by approximately 30% thanks to large cyclic load fluctuations characteristic of IPTV delivery. The presented offline algorithm can be used as a tool to benchmark new DPM-based online algorithms. We intend to release our code as free and open-source software.