Recent collapses of SIP servers in the carrier networks indicates two potential problems of SIP: (1) the current SIP design does not easily scale up to large network sizes, and (2) the built-in SIP overload control mechanism cannot handle overload conditions effectively. In order to help carriers prevent widespread SIP network failure effectively, this chapter presents a systematic investigation of current state-of-the-art overload control algorithms. To achieve this goal, this chapter first reviews two basic mechanisms of SIP, and summarizes numerous experiment results reported in the literatures which demonstrate the impact of overload on SIP networks. After surveying the approaches for modeling the dynamic behaviour of SIP networks experiencing overload, the chapter presents a comparison and assessment of different types of SIP overload control solutions. Finally it outlines some research opportunities for managing SIP overload control.
TopIntroduction
Internet telephony is experiencing rapidly growing deployment due to its lower-cost telecommunications solutions for both consumer and business services. Session Initiation Protocol (SIP) (Rosenberg et al., 2002) has become the main signaling protocol to manage multimedia sessions for numerous Internet telephony applications such as Voice-over-Internet Protocol (IP), instant messaging and video conferencing. 3rd Generation Partnership Project (3GPP) has adopted SIP as the basis of its IP Multimedia Subsystem (IMS) architecture (3GPP WG, 2011). With the 3G (3rd Generation) wireless technology being adopted by more and more carriers, most cellular phones and other mobile devices are starting to use or are in the process of supporting SIP for multimedia session establishment (Faccin, Lalwaney, & Patil, 2004).
Figure 1 illustrates a simplified architecture of a SIP network. A SIP network consists of two types of basic elements: User Agent (UA) and Proxy Server (P-Server) (Rosenberg et al., 2002). A user agent can act as a user agent client (UAC) or as a user agent server (UAS). A P-server not only acts as the contact point with UA for core network service access but also provides routing for the signaling messages. SIP is responsible for establishing, modifying and terminating sessions for multimedia communication among multiple UAs.
Figure 1. Simplified architecture of a SIP network
RFC 5390 (Rosenberg, 2008) identifies the various reasons that may cause server overload in a SIP network. These include but are not limited to poor capacity planning, dependency failures, component failures, avalanche restart, flash crowds, denial of service attacks, etc. In general, anything that may trigger a demand burst or a server slowdown can cause server overload and lead to overload propagation and server crash, thus bringing down the whole SIP network.
The objective of this chapter is to present a systematic investigation of current state-of-the-art SIP overload control algorithms which aim at preventing server crashes in carrier networks. In order to provide a better knowledge of the major cause of SIP network collapse, the next section reviews two basic mechanisms of SIP, and describes the existing works on the performance study of the SIP overload. The third section surveys the related SIP modeling and analysis which can help network planners, operators, and researchers to understand how server overloading and widespread SIP network failure may happen under short-term demand bursts or server slowdowns. The forth section makes a comparative study of different types of SIP overload control solutions, thus helping carriers choose the appropriate solutions to avoid potential SIP network collapse (e.g., Skype outage (Ando, 2010) or VoIP outages in British Telecom, Vonage and Wanadoo (Materna, 2006)) in different overload situations. Finally, some future works for the SIP overload control are discussed.
TopSip Overview
SIP works in the application-layer for session establishment. Figure 2 depicts a typical procedure of a session establishment. To set up a call, a UAC sends an “Invite” request to a UAS via the two proxy servers. The proxy server or the UAS returns a provisional “100Trying” response to confirm the receipt of the “Invite” request. The UAS returns an “180Ring” response after confirming that the parameters are appropriate. It also evicts a “200OK” message to answer the call. The UAC sends an “ACK” response to the UAS after receiving the “200OK” message. Finally the call session is established and the media communication is created between the UAC and the UAS through the SIP session. The “Bye” request is generated to finish the session, thus terminating the communication.
Figure 2. A typical procedure of session establishment