A simulation-based performance evaluation of Tahoe, Reno and Sack TCP as appropriate transport protocols for SIP

摘要

The Session Initiation Protocol currently supports the use of both the Transmission Control Protocol and the User Datagram Protocol as transport protocols. Proponents of the SIP protocol have cited this characteristic as a laudable consequence of the modular nature of SIP. However, the explicit support of an unreliable protocol such as UDP leads to, arguably, unnecessary complication in the protocol standard. The matter has been the subject of recent debate with some favouring the deprecation of explicit support for UDP. This paper seeks to compare the performance of different TCP variants, specifically the Tahoe, Reno and Selective Acknowledgement (Sack) variants in the particular context of SIP signalling. We simulate a scenario that is likely to receive the maximum benefit from TCP's end system-based traffic control functionality; we present results and draw conclusions from our evaluation.