Modeling and performance analysis of resource allocation strategies for real-time services in UMTS using TIPPtool

Abstract

The third generation UMTS system will be able to offer a wide variety of multimedia services (speech, teleconferencing, browsing on the net, etc.). In order to allocate resources to these heterogeneous traffic sources in an efficient way, shared channels have been introduced in addition to the traditional transfer on dedicated channels. Although initially conceived for not real-time packet users, shared channels can satisfy the performance requirements of real-time traffic, thanks to the short and efficient channel reservation signaling phase. This paper compares the performance of the radio resource allocation strategies based on dedicated and shared channels, and quantifies the increase in system capacity of an hybrid solution, where shared channels are used to convey the bursty low priority component of a real-time video service. Since the low priority packets that cannot be delivered are dropped, a specific minimum amount of shared resources is required to match the global quality of service requirements of the video service.

Modeling and numerical analysis are carried out with the TIPPtool package. With its formal and effective notation and the provision of stochastic features, a compact system specification can be obtained and validated for proper behavior, and performance measures can be derived. In order to numerically evaluate the huge state-space of the complex system under study, a compositional approach is proposed: the system is decomposed into simpler components, which are analyzed separately and whose numerical results are associated for global performance evaluation.

Introduction

The third generation (3G) of mobile telecommunication networks embraces some international standards, currently still under definition, which must comply with the general requirements of the 3G global system International Mobile Telecommunications 2000 (IMT-2000): they must support a wide variety of services homogeneous to fixed networks’ ones, which must be available always (on any terminal the users have at disposal, due to the personality of the connection) and everywhere (in any region of the globe, by means of satellite as well as terrestrial coverage), with a bit-rate varying from 144  kbps to 2 Mbps depending on the mobility profile of the user.

Universal Mobile Telecommunication System (UMTS), the European proposal for the 3G mobile system, has been conceived to support multimedia services (audio and video on Internet, video-telephone and video-conferencing, etc.) in a flexible and efficient way. UMTS will enable them by satisfying their larger bandwidth requirements as well as their specific quality of service (QoS) demands [1]. A good overview on concepts of 3G mobile systems and on the characteristics of UMTS air interface can be found in [2].

The higher capacity is, actually, one of the main features of the 3G wireless networks: unlike the 2G Global System for Mobile (GSM) communications, which is based on the division of the spectrum in narrowband radio carriers, UMTS is a wideband system, where the maximum available bit-rate per physical channel is half the bit-rate of the spread-spectrum signal (3.84 Mchip/s), thus largely improving the about 160 kbps achievable by GSM and General Packet Radio Service (GPRS) in their most sophisticated hardware implementations (Phase 2 plus of GSM standard).

The second issue concerns the QoS parameters (average and peak bit-rate, residual bit error rate, transmission delay, etc.) that the network must guarantee, within a certain degree, to each data transfer originated in the system. Multimedia services comprise more than one media component with peculiar characteristics: for efficient radio resource usage, they can be conveyed on separate radio connections, each fitting the specific QoS requirements. Besides, according to source coding theory, even within a single data flow (e.g. the audio component of a videoconference) it is possible to discern groups of bits that show a different sensitivity to errors and thus require differentiated error protection (for speech multimode source encoding see [3]). A data stream can be divided into several components that affect the quality of the received signal in a different way, i.e. the more relevant their contribution to the perceived signal quality is, the more stringent their QoS requirements are.

UMTS supports a multiplicity of connections associated with a single call, each of which is established on a variable bit-rate radio bearer that relies on dedicated physical channels (DCHs) in case of circuit-switched connections. For more bursty data flows, UMTS provides a connection less data transmission, which originally relied on the same physical channels used for mobile-originated access attempts and network replies (Random Access Channel and Forward Access Channel). More recently, to overcome the scarce efficiency of the RACH/FACH solution at high traffic loads, a set of uplink and downlink shared physical channels has been introduced and discussed within Third Generation Partnership Project (3GPP), the technical body responsible for UMTS standardization. Shared channels (SCHs) are reserved to discontinuous traffic sources and assigned on demand following a light signaling protocol: this enables minimizing channel acquisition delays and satisfying the severe time constraints of several real-time applications. In fact, although SCHs have been designed for usage by time-unconstrained packet-switched services based on TCP/IP protocols, it is also interesting to evaluate how much variable bit-rate real-time services can profit from them. In this work, the performance of the SCH solution and the traditional data transfer on DCHs are compared in the hypothesis of multiclass mobile-originated traffic of speech and video. Video sequences are encoded in a scalable way, i.e. consist of high priority and low priority binary streams, modeled with a more general approach than in [4], where preliminary results were presented.

One of the main techniques used in the performance analysis of mobile systems and radio access protocols is the theoretical approach based on queuing theory and traffic analysis: in the wireless scenario under study, this method is strongly complicated by the differentiated QoS requirements of the two classes of service (in the hypothesis of single service traffic load this technique is less difficult to apply). An alternative approach is to build a simulator of the radio interface, either implemented in a programming language, or by means of existing network simulators, which require configuration and specification files written in an appropriate meta-language. In both cases, new call arrivals and service statistics require a multidimensional birth–death process and lead to models that, despite their high detail, can be tedious to debug and time consuming to simulate. In addition, it is almost impossible to know if the simulated system corresponds exactly to the protocol under study: large models can be error-prone, particularly when the communication between system entities is ruled by complex signaling protocols. Moreover, measure results are always affected by the quality of random generators and the choice of the simulation time.

To overcome these difficulties, in this paper an approach based on a stochastic process algebra formalism is adopted to study the behavior and the performance of DCH and SCH resource allocation strategies. Modeling and performance analysis are carried out with the TIme Processes and Performance evaluation tool (TIPPtool) package [5]. Giving emphasis to compositionality, concurrency and communication issues as well as quantitative aspects, TIPPtool provides a self-contained framework where the protocols are described, verified for proper behavior and analyzed for performance evaluation. TIPPtool supports a theoretical approach to system performance measures: in the formal specification phase a compact high-level stochastic model of the communication system can be built and tested; in the numerical evaluation phase the prototype’s state-space is naturally converted into a Markov chain (MC), the model of reference in queuing theory.

The rather complex systems under study lead to enormous state-spaces for which efficient reduction functions are not yet available within the tool. Therefore, several approximations are necessary: the system is decomposed into simpler components, which are solved in a compositional way and whose results are associated to give global performance evaluation. Differently from [4], in this paper a more accurate solution is obtained by applying an iterative approach to the performance analysis of the transmission on dedicated channels. In order to validate the impact of these approximations, the performance measures calculated by TIPPtool are compared with simulation results obtained with Telecommunications Description (TeD) language [6], a simulation tool devoted to telecommunication networks, which renders the events’ statistics in a very accurate way but, lacking in model-checking facilities, requires particular carefulness and expertise in testing communication and synchronization aspects.

This paper is organized as follows. Section 2 contains an overview of TIPPtool syntax, operational semantic and model-checking as well as performance evaluation capabilities. Section 3 illustrates the case study: after presenting the technical features of UMTS air interface and the working assumptions, the resulting TIPPtool models and the compositional modeling strategy adopted to limit the state-space are described in detail. Numerical results are presented in Section 4 and conclusive remarks and future work are summarized in Section 5.