Evaluating Performance Characteristics of SIP over IPv6 Thomas Hoeher 1 , Martin Petraschek 1 , Slobodanka Tomic 1 , and Michael Hirschbichler 2 1 ftw. (Telecommunications Research Center Vienna), Vienna, Austria Email: {hoeher, petraschek, tomic}@ftw.at 2 Institute of Broadband Communications, Vienna University of Technology, Vienna, Austria Email: michael.hirschbichler@tuwien.ac.at Abstract— Due to the ongoing massive growth of the global Internet, the rising integration of Voice over IP (VoIP) services and the Fixed Mobile Convergence (FMC), the IPv6 protocol and the Session Initiation Protocol (SIP) are key technologies for the realization of next generation communications. For both topics, IPv6 and SIP, a lot of self-contained research has been done. However, the challenge of SIP over IPv6 as well as related issues and performance impacts were not considered so far. In this article, we close this gap and draw attention to theoretical and practical aspects of the integration of SIP and IPv6, referred to as SIPv6. In this context our special interest concerns the interworking of heterogeneous IP networks during the transition from IPv4 to IPv6 and their ramifications on the VoIP service. Inevitably, during this period of co-existence the available transition techniques have an impact on the network and application performance. To quantify this impact, we set up a SIPv6 VoIP testbed and measured the performance penalties introduced by four selected transition techniques. We characterize the performance of transition scenarios compared to native scenarios by presenting measurement results and gained insights. Our study reveals individual pros and cons of transition technologies and their available implementations. Index Terms—IPv6, SIP, performance measurements, prox- ying, tunneling, transition techniques, 6to4, Teredo I. I NTRODUCTION More than one decade ago the Internet Engineering Task Force (IETF) has started to develop a successor for the most widely deployed network layer protocol in the Internet: the Internet Protocol version 4 (IPv4). IPv4 was originally developed in 1981 to solve the (internet) routing matters for a small backbone connecting academic and government networks within the United States. Nowadays, the Internet is a worldwide backbone interconnecting thousands of autonomous systems, how- ever the network layer protocol is still IPv4. As nobody could have foreseen this fulminant growth, the developers of IPv4 decided generously to serve almost This contribution is based on “Performance Evaluation of SIPv6 Tran- sitioning,” by T. Hoeher, M. Petraschek, S. Tomic and M. Hirschbichler, which appeared in the Proceedings of The Second International Multi- Conference on Computing in Global Information Technology (ICCG 2007) co-located with IPv6TD 2007: The Second International Work- shop on IPv6 Today - Technology and Deployment, Guadeloupe, French Caribbean, March 2007. c  2007 IEEE. This work was supported by the Kplus program of the Austrian Federal Government. four billions of nodes, i.e., half of the today’s world population. Nevertheless, for some years now the Internet is on the verge of depletion of IPv4 addresses. Already in 1992 the IETF identified this upcoming exhaustion and started to specify countermeasures such as Network Address Translation (NAT), Variable Length Subnet Mask (VLSM), Classless Interdomain Routing (CIDR), and Internet Protocol version 6 (IPv6). However, the only sustainable solution to cope with IPv4 address space shortage, which does not solely shift the point of depletion is IPv6. IPv6 was conceived within the IETF working group IPng founded in 1994 with the goal to define the next generation Internet Protocol. Among several proposed alternatives, IPv6 [1] - originally called IPng - has been selected as the successor of IPv4. Today, in a variety of IPv6-related RFCs, complementary topics like security, address and routing schemes, the IPv6 transition, and mobility enhancements are treated. In general, IPv6 offers some novelties and benefits but still the most convincing argument for introduction is the 128-bit address space. The urgency for the introduction of IPv6 becomes more and more obvious since well known Internet engineers like Tony Hain (Cisco Inc.) and Geoff Huston (APNIC) predict the point of depletion between autumn 2008 [2] and summer 2012 [3]. In other words, it is high time to continuously push the global implementation of IPv6. One of the still missing pieces in the IPv6 puzzle is the evaluation of typical deployment issues such as per- formance, interoperability, and scalability. In this context, the main criteria at the beginning of IPv6 introduction would be the seamless interworking between IPv4 and IPv6, also considered as IPv6 transition. Apparently, the migration from IPv4 to IPv6 could only happen in the course of an incremental transition where both protocols have to co-exist. The duration of this process is not predictable, however it brings up a variety of additional aspects, such as performance and scalability in particular at the application layer. These open questions motivated our IPv6 research on the transition aspects of SIP (Session Initiation Proto- col) [4], the protocol which is currently penetrating and revolutionizing the Internet and its services landscape. But not only the Internet, it is about to change the entire 40 JOURNAL OF NETWORKS, VOL. 2, NO. 4, AUGUST 2007 © 2007 ACADEMY PUBLISHER