以服務為導向之架構提供了一個具擴充性、跨平台的方法來組織分散式的資源。其架構最廣泛的實作便是網路服務（ Web services ），而網路服務的必要功能之一就是網路服務搜尋（ Web services discovery ）。目前集中式的網路服務搜尋架構如 Universal Description and Discovery Integration (UDDI) 已知有單點故障與擴充性之問題。然而，現有點對點架構：以Gnutella為基礎或以 Chord為基礎之架構，則因需做多點跳躍才能取得所需的資料，在無線網路上會造成延遲增加之現象。為了解決單點失敗、擴充性、延遲之問題，本論文於行動網路中提升 Web 服務探索之效能。此系統提出點對點 Web 服務探索，藉由 Pastry 系統建構點對點網路，並採用 Pastry 之繞徑演算法及資料結構，提供分散且具減少延遲之語意搜尋服務，實現了 Web 服務在同儕網路上以關鍵字為基礎之服務搜尋。根據模擬與實驗數據顯示，延遲比起以 Chord 為基礎之服務搜尋改善了百分之六十七，並且能夠平衡負載，解決單點失敗之問題。 Service-Oriented Architecture (SOA) provides a scalable and interoperable paradigm for organizing distributed resources. In general, the SOA is implemented by Web services, and discovery service is the fundamental process of services engagement in Web services. However, a centralized architecture for Web services discovery, such as Universal Description and Discovery Integration (UDDI), goes through the single node of failure and scalability problem. Otherwise, the existing peer-to-peer architectures, such as Gnutella-based or Distributed Hash Table-based, were suffered from larger delay due to multi hops delivery in mobile networks especially. In this thesis, a Performance Improvement in Web Services Discovery for Mobile Networks is proposed, and the architecture is a decentralized and adaptive service discovery with enhancement of matching capability. This architecture adopts a distributed index system with support of Pastry overlay networks for services discovery. The keyword query for semantic search in peer-to-peer network is supported in the proposed architecture as well. Furthermore, the routing procedure and leaf set of Pastry is adopted to reduce delay in mobile networks. Results from simulation show that our proposed solution is scalable and improved the performance up to 67.85% compare to Chord-based solutions in mobile networks.