22 VOLUME LXIV. • 2009/II 1. Introduction One of the most important time critical services of cur- rent and also of ITU-T NGN (Next Generation Network) communication networks in the near future is voice traf- fic. The VoIP(Voice over IP) technology radically trans- formed also the cost of telephone service and the beha- vior of subscribers. IP phone service exploits efficiently the Internet based network infrastructure and approxi- mates the quality of PSTN traffic services. The best-ef- fort transmission mechanism of IP networks cannot as- sure guarantees for delay sensitive voice traffic, hence for successful utilization of VoIP QoS techniques are needed among the end nodes. The evaluation charac- teristics of modeling aggregated voice and other type of traffics implies selection of optimal QoS mechanisms. The network traffic coming from a voice source depends strongly on the utilized voice codec type. These codecs are grouped into two classes: constant bit rate mechanisms (e.g. G.711), as well as si- lence supression mechanisms based on re- peating ON and OFF periods of activity (e.g. G.728, GSMFR, G.722) [4]. Actual packet switching voice systems in- clude not only IP traffic capable phone end nodes, but application servers responsible for signaling and cost accounting as well (Fi- gure 1). Signaling methods (like SSCP, SIP, etc.) in this environment are much more intelligent than those used in PSTN networks (e.g. QSIG). The voice content transmission is realized by RTP (Real Time Protocol) directly between IP phone nodes. Signaling is transmitted in TCP, and digitized voice is transmitted in UDP seg- ments. To interconnect VoIPand PSTN networks special gate- ways are used that are capable of converting signaling and voice content as well. The IP phone compiles mes- sages from the sampled voice and assembles voice seg- ments by a codec module (Figure 2). Voice segments transmitted by RTP on top of UDP are shaped by jitter buffers at the receiving node. Modules like G.711, G.723, G.728, and GSM are called narrow-band codecs and uti- lize at most 64 kbps voice rate [4]. Codecs like Brand- Voice32, G.722, etc. need higher bandwidth for the in- creased voice quality. The main characteristics of IP voice codes are: voice bit rate, length of voice frame (80-520 bytes), duration of voice frame (0.125-20 ms), the IP packet bandwidth (24- 272 kbps) and the delay of voice transfer (0.25-40 ms). Figure 1. IP phone and VoIP architecture Keywords: NGN, TCP, UDP, codec, QoS, DiffServ, self similarity, wavelet, fractal, entropy Strict requirement is emphasized regarding QoS guarantees of the NGN (Next Generation Network) networks today. DiffServ mechanism is applied mostly for classification of protocol data units of real time and conventional information streams in LAN/MAN environment. The dependence of VoIP traffic characteristics of the delay and the jitter sensitive IP telephony vs. voice codec applied can be considered an exciting scientific question. We analyze Ethernet traffic generated by G.711, G.723, G.728 and Wideband (G.722) voice codecs. The self similar, fractal and multifractal properties of popular TCP based services (http, ftp, telnet, etc.) in LAN/MAN environment are well known for several decades. In this paper, we study the effect of UDP based current voice mechanisms on self similarity of the Ethernet data traffic. UDP traffics of the IP phones are evaluated in congested and congestionless environment using sophisticated methods of entropy and wavelet analysis. A new and efficient evaluation method, named ON/(ON+OFF) transformation is applied to the characterization of VoIP traffic.  RESEARCH VoIP LAN/MAN traffic analysis for NGN QoS management ZOLTÁN GÁL Center for Information Technologies Centre of Arts, Humanities and Sciences, University of Debrecen zgal@unideb.hu