I. Introduction
As an indispensable building part of Internet of Things (loTs), wireless sensor network (WSN) has been widely used in many applications, including smart home, e-health and environment monitoring. Most existing large-scale WSNs follow a tiered architecture [1] as shown in Fig. 1, which consists of a large amount of resource-limited sensor nodes in the lower tier and several resource-rich storage nodes in the upper tier. Sensor nodes are mainly responsible for sensing, while storage nodes store data received from sensor nodes and answer queries received from the sink. There are several advantages of the tiered WSNs. On the one hand, sensor nodes can save much energy and avoid communication bottleneck by sending data to their closest storage nodes instead of the sink [2], meanwhile, it can also reduce storage cost. On the other hand, the query processing becomes more efficient since the sink only needs to communicate with several storage nodes rather than a large amount of sensor nodes [3]. Therefore, the tiered architecture is favorable for extending capacity and improving scalability of WSNs.