Teletraffic Analysis of OFDMA Cellular Systems With Persistent VoIP Users and Maximum SIR Scheduling Based on Order Statistics

Abstract

Voice over IP (VoIP) is the multimedia service with the fastest growth on current cellular systems and this trend will continue in the future. Voice transmission by VoIP users with traditional dynamic scheduling schemes could become a bottleneck in the signaling channels due to the periodic transmission of signaling small packets. To solve this problem, persistent scheduling has been proposed. Since the resource allocation information is indicated in the first burst of persistent data, no information resource allocation is required for the duration of the data bursts. This paper presents a performance study of the persistent scheduling based on the transmission of VoIP traffic, using as mathematical tools order statistics and teletraffic analysis. A combination of interference analysis and birth and death processes is performed to obtain the average transmission rate in the system using the Shannon equation. We obtain the order statistics of band signal-to-interference ratios (SIRs) by considering the co-channel interference. The order statistics are then used to analyze a best-m band feedback scheme, in which every user reports the index of their best-m bands to the base station. If multiple users present a band as their best band, two different approaches are analyzed: 1) random allocation and 2) Maximum SIR allocation. The probability density function of the SIR was obtained; it is observed that the Maximum SIR allocation has a significant gain of up to 20% in regard to the random allocation. After that, the probability of using a specific modulation and coding scheme is obtained. In order to obtain the probability of forced termination and the average data rate, these probabilities are introduced in the birth and death process as a different arrival rates and service times. It is observed that the order statistics model the cellular systems are more accurate. As an example for uniform user distribution and voice over VoIP traffic, the numerical results shown that the probability of forced termination is corrected by more than 70% in comparison without order statistics. Moreover, using the methodology developed with order statistics, an evaluation is made for a non-uniform user distribution with Skype traffic, considered as a very important service for 5G cellular systems.