Since improving energy efficiency (EE) must not sacrifice quality-of-service (QoS), the EE-QoS relation is a fundamental issue for EE optimization. This paper provides an overview of our research on the EE-QoS relation and its application to elastic access optimization in hyper-cellular networks over the past five years. Because delay bound is a representative QoS requirement, we first discuss our findings on the EE-delay relation, and show that if average total power consumption is linear with average service rate, then there exists a non-tradeoff region in the EE-delay relationship. We then summarize how to design energy efficient resource allocation for different kinds of services, such as real-time and non-real-time services, and ultra-reliable and ultra-low latency service. For real-time service, in order to achieve the optimal EE-delay relation, resource allocation should depend on queue length. For delay tolerant service, it is possible to leverage predictive information for making a resource allocation plan and pushing data. The results show that the trajectories and preferences of users are very useful for improving EE in wireless systems. Finally, we summarize preliminary results for the resources required to maximize EE under the constraints of ultra-reliable and low-latency network.
(Color online) Typical timescales of wireless channel and delay requirements of different services
(Color online) Impact of large scale channel gain on power-rate relation