以SIP為基礎的行動管理之研究

Abstract

隨著日益普及的多功能行動裝置以及搭載於其上各式各樣推陳出新的即時應用與服務之快速發展，無縫的終端移動支援顯然成為行動管理當中最重要的需求。從跨網域移動行為的角度來觀察，無線網路環境底下的通話初始協定（Session Initiation Protocol，以下簡稱SIP）存在了封包遺失和通話失敗兩項重要的議題，嚴重地造成通話的服務品質下降。然而，這兩大關鍵議題分別起因於以下三種問題的發生：（1）通話過程中發生移動；（2）所歸屬之網域行動資料庫發生異常；（3）通話之前發生移動。在這篇論文中，我們提出了一個系統化的行動解決方法，藉由移動用戶行為探勘所萃取出來的移動樣式以達到移動快速（延遲時間縮短）與平滑（通訊維持流暢）的目的。我們將此方法命名為STAMP。在本文中，移動樣式被賦予IP網路封包交換的特性，使移動樣式能夠適時地輔助封包亦或通話請求訊息轉送至使用者目前所在的位置。經由7個模組所建構而成的4個階段，”STAMP”將透過兼具效率與效力的序列樣式探勘過程，依循各個問題專屬的最佳路徑，適性化地解決上述的三種問題。最重要的是，我們所設計的方法提升了SIP即時服務的品質。另外，更進一步將我們的方法STAMP在通訊信令所發生延遲、中斷的時間與相關文獻所提出來的方法Shadow Registration做比較與分析。

With rapid development of widespread multi-function mobile devices and a variety of real-time services, seamless terminal mobility support becomes the most significant requirement in mobility management. Based on Session Initiation Protocol (SIP) over Wireless LAN, there exist two critical issues from the perspective of interdomain handoff, including packet loss and call failure, greatly diminishing the quality of service. These issues arise from Mid-Call Mobility Problem, Home Registrar Failure Restoration Problem, and Pre-Call Mobility Problem, respectively. In this article, we propose a systematic mobility scheme, “STAMP”, to achieve not only fast handoff but also smooth handoff, by means of extracted knowledge from mobile user behavior mining. The moving pattern here is endowed with the nature of IP network as an assistant to forward packets timely to the right place. Through 4 phases composed of 7 modules, each of the three problems is solved adaptively via an exclusive path in “STAMP” with effective and efficient sequential pattern mining processes. Above all, we enhance the quality of real-time SIP services. We also analytically compare the delay and disruption time for signaling associated with Shadow Registration and our solution.