超立方體(Hypercube)結構近年來已被廣泛的應用在許多網路系統當中，其中應用在具有位置知覺(Locality-Aware)及興趣分群(Interest-Based)系統的同儕網路(Peer-to-Peer ; P2P)架構下，超級結點(super node)為底層結點與外界間之通訊、傳輸等橋梁。因此如何有效率的與底層結點作溝通聯繫，將攸關著此網路系統之效率。本篇將提出利用改良式最小成本生成樹演算法，來達成超級結點與其底層結點間之搜尋時間更為快速之目的。 Hypercube structures have been used in many networks in the past few years. One of these systems is locality-aware hybrid system with interest grouping. In this system, super nodes have more resources and bandwidth. Intra-peers and inter-peers are separated by super node. When intra-peers need to communicate with inter-peers, they will send messages by way of the super node. It is important to minimize the transmission time between intra-peers and the super node. In this paper, we proposed an algorithm based on the minimum cost spanning tree. Our purpose is to shorten the transmission time from the super node to intra-peers.