Performance Modeling of Wireless Voice over IP

Abstract

The recent popularity of voice over IP (VoIP) services in the Internet world lies in the fact that in VoIP, voice packets are carried over the shared packet-switched network (the Internet) rather than on dedicated circuit-switched telephony network. This phenomenon reduces the cost of carrying the voice packets substantially. We propose that the concept of VoIP over wireless GPRS (General Packet Radio Service) also reduces the complexity of the GSM/GPRS architecture. Currently, the existing GSM (Global System for Mobile Communications) network carries the voice packets in the air link and converts it to circuit-switched PCM (Pulse Code Modulation) signal by using the voice transcoder. A 64 Kbps circuit-switched connection is then established through the existing PSTN (Public Switch Telephone Network). Wireless data packets on the other hand, share the same physical air link, but are carried through the GPRS packet-switched network. If we carry the voice on the same packet-switched network rather than on a separate circuit-switched network, then integration of data and voice would alleviate the necessity of circuit-switched components. However, maintaining the voice quality offers significant challenges. In this paper we propose a method of implementing the VoIP in GPRS network using H.323 protocol, and also identify the additional architecture requirements that will address some of the performance impediments of current GPRS VoIP implementation. The performance results for control message signals demonstrate the possibility of maintaining call set-up delay within the current delay standards. Analytical modeling of the voice traffic shows that significant improvements in terms of voice traffic capacity is possible with the help of a silent-detection scheme.

This work is partially supported by NSF grants under EIA-0086260, EIA-0115885, IIS-0121297 and Texas Telecommunication Engineering Consortium (TxTEC).