實現SIP高可用性網路之快速錯誤偵測與復原機制

Abstract

SIP代理伺服器在SIP網路中是一個重要建構單元，它可以組成一個叢集來避免軟體或硬體錯誤所造成的服務中斷。此外搭配一個分派器的SIP代理伺服器叢集設計仍然會遭遇到單點失敗的問題。現有的研究曾提出使用雙SIP分派器的架構，並且在IP接手機制中使用VRRP (Virtual Router Redundancy Protocol)。在這篇論文中，我們設計且實作了一個可靠的SIP網路，它包含了 n + k 個SIP分派器 (n 個使用中的分派器 + k個備援的分派器) 去控制並監測提供建立VoIP與視訊會議連線的m個SIP代理伺服器所組成的叢集。我們提出一個快速錯誤偵測與復原(FFF)機制。FFF使用OpenAIS作為一個高可用性中繼軟體去實現分派器與代理伺服器的健康偵測與錯誤復原。實作量測結果顯示，FFF機制比VRRP機制減少了百分之八十之的錯誤復原時間，以及比OpenSER機制減少了百分之八十五的錯誤復原時間。

The SIP proxy server, which is an important building block of a SIP network, can be in a form of a cluster to prevent service unavailability caused by software or hardware failures. The design of a SIP proxy server cluster with a single dispatcher still faces the single-point-of-failure problem. The use of dual SIP dispatchers architecture was proposed and VRRP (Virtual Router Redundancy Protocol) was used as an IP failover mechanism. In this thesis, we have designed and implemented a dependable SIP network that includes n + k SIP dispatchers (n active dispatchers + k backup dispatchers) to control and monitor a cluster of m SIP proxy servers for VoIP and Video conferencing applications. A fast failure detection and failover (FFF) scheme has also been proposed. FFF uses OpenAIS as a high availability middleware to perform health check and failover of dispatchers and proxy servers. Experimental results have shown that the FFF scheme reduces the dispatcher failover time by 80% compared to the VRRP mechanism and also shorten the proxy failover time by 85% compared to the mechanism provided by OpenSER itself.