容錯ICA設計

Abstract

隨者網際網路的蓬勃發展，網際網路語音電話已經發展到可以用來取代傳統電話。網際網路語音可以提供比傳統電話更多的優點，比如像建構成本低廉、電話費率低廉、可提供語音和數據的整合應用、以及系統業者易於開發新服務。目前的網際網路語音協定主要由International Telecommunication Union (ITU)和Internet Engineering Task Force (IETF)所制定。ITU是最早提出網際網路語音的協定，所提出的是H.323；而IETF先後分別提出的Media Gateway Control Protocol (MGCP)、和Session Initiation Protocol (SIP)。雖然以上協定已經逐漸發展成熟，但在傳統的電信網路宣稱有99.999%的高可靠度，因此若要取代傳統電話，網際網路語音服務的可靠度成了重要議題。 我們使用一台具有監聽功能的備用伺服器，來增加網際網路語音服務的可靠度。要提供高可靠度的服務，必須要考量到硬體、網路、伺服器等等的狀況。隨者硬體的考靠度增加和使用多重的硬體，將可以減少硬體、和網路錯誤的發生。因此，伺服器的可靠度變得非常的重要。由於網際網路語音具有及時性和每一通電話有它自己的狀態，因此我們的備用伺服器能夠監聽發送給主要伺服器的封包，維護和主要伺服器一樣通話狀態和通話資料。因此，當主要伺服器失效時，備用伺服器能夠偵測到，並且在短時間內提供服務而不需要執行耗時的復原程序。

With the rapid development of the Internet technologies, Voice over IP (VoIP) telephony may replace traditional telephony. VoIP telephony has more advantages than traditional telephony, such as lower cost to set up the VoIP system, a lower billing rate for VoIP users, the integration of voice and data applications, and an easier way to develop new services for system operators. Current VoIP standards are proposed by International Telecommunication Union (ITU) and Internet Engineering Task Force (IETF). ITU proposed H.323, and then IETF proposed Media Gateway Control Protocol (MGCP), and Session Initiation Protocol (SIP). Since the public switched telephone network (PSTN) system has five nines reliability (99.999%), it is critical to design a highly reliable VoIP system. We use a hot standby redundant server with a monitor function to enhance the availability of VoIP systems. For a highly available system, we must consider hardware, network, and server failures. With the increase of hardware reliability and using redundant hardware to set up the system, hardware, and network failures can be reduced. Thus, server failures become the most important issue to deal with. VoIP applications are real-time services and each call has it own call data and call status maintained by a call server. In our system, the standby server can monitor the packets sent and received by the active server, and process the same call flow as the active server to maintain the call data and call statuses. When detecting failure of the active server, the standby server can take over with a short recovery routine.