We consider optimal energy-aware load balancing of elastic downlink data traffic inside a macrocell with multiple small cells within its coverage area. The system is modeled as a set of parallel queues. In particular, the model of the small cell includes the setup delay resulting from activating the base station after being placed in a low power off state and the idle timer controlling the amount of time to wait before being switched off. We apply the theory of MDPs to develop state-dependent dynamic policies for controlling both the routing of the arrivals as well as the length of the idle timer that minimizes the weighted sum of energy and performance. In particular, we show that in the optimal policy the idle timer control can be simplified to selecting a value arbitrarily close to zero or infinite. Additionally, by utilizing the first step of the well-known policy iteration method, we develop an explicit near-optimal dynamic policy for routing the arrivals and also for determining the idle timer configuration of the system, based on the expressions for the future marginal costs. The performance of the policy is illustrated through numerical examples.
|Number of pages||13|
|Publication status||Accepted/In press - 2019|
|MoE publication type||A1 Journal article-refereed|
- HetNets, load balancing, performance-energy tradeoff, parallel queues, Markov Decision Processes