P2P network is a focus research field with rapid development in recent years. Distributed hash tables (DHT for short) schemes are the core technique to locate and search resources in structured P2P network. However, due to the characteristic of DHT that its hash method destroy data locality, DHT only support exact-match search, cannot support complex search process. A simple and flexible solution has been built to tackle the problem which is known as over-DHT paradigm without need to modify existing DHT infrastructures which significantly decreases the complexity of deployment. Among the well known over-DHT paradigms now, LIGHT is an elegant one. It provides more efficient complex query methods than other paradigms and also reduces the paradigm maintenance cost. However, LIGHT does not make full use of the advantages of space partition tree such as that it is a complete binary tree and divides data space into uniform interval in each level of the tree. In this paper, we propose an improved range query algorithm, Upper Boundary Predictably Lookup (UBPL for short) to reduce the DHT-lookup times during range query procedure in LIGHT. We also proved the effectiveness of UBPL by simulating experiments.