内容发布更新时间 : 2024/11/20 18:43:15星期一 下面是文章的全部内容请认真阅读。

6.结语

本文提出了一个通过GSM-SMS移动浏览的服务质量评价的初步研究WAP的方案。一个城市的旅游服务已被设想。因此，一个合适的网络体系结构已被提出。模型已被定义为特征的下行流量通过WAP浏览用户产生。在甲板上的卡字节的长度已统计特征的考虑不同于WAP服务器发现很多例子。一个分析方法已被描述，允许预测的平均甲板延迟作为一个函数的同时参与WAP会话的用户数。此结果允许估计的服务质量向用户提供有益的思考，可以得到设计的WAP服务 参考 [1] S.南大，K. Balachandran，S.库马尔，“在无线分组数据技术的适应服务”，IEEE通信杂志，第54-64页，2000年1月。

[2] http://www.wapforum.org/what/WAP_white_pages.pdf

[3] S.M.雷德尔，M?。韦伯，M·W·奥利芬特。简介GSM,艺达大厦，MA，1995年 [4] ETSI：“数字蜂窝通信系统（阶段2+）;点对点（PP）短消息服务在移动无线电接口（SMS）支持”（GSM 04.11）。

[5]无线应用协议论坛有限公司，WAP-158，无线数据报协议规格，版本号1999年11月5日。 [6] ETSI - 数字蜂窝通信系统（阶段2+）;技术实现的短消息服务（SMS）;点对点（PP）（GSM 03.40）。

[7]无线应用协议论坛有限公司，WAP-154，二进制XML内容格式规范1.2版，1999年11月4日。

[8] ETSI - 数字蜂窝通信系统（阶段2+）;信道编码，（GSM 05.03版本5.5.0） [9] A. E.品牌，A. H. Aghvami，“多维PRMA与广播优先级的贝叶斯 -

一个MAC策略多业务流量在UMTS”，硕士论文高科技第47期，4号，第1148至1161年，1998年11月。

Performance Evaluation of the WAP Protocol over

SMS in a GSM Network

Alessandro Andreadis, Giuliano Benelli, Giovanni Giambene, Bernardo Marzucchi,

1

1

1

1

David Sennati

1

1

Dipartimento di Ingegneria dell’Informazione - Università degli Studi di Siena, Via Roma, 56

- 53100 Siena, Italy

Abstract: The widespread diffusion of mobile communication systems pushes research and development efforts towards the implementation of new services accessible through mobile terminals. This work has been carried out in the framework of the PALIO Project in the fifth Research Framework of the European Commission. It has been considered a tourist service where each user may access information directly through its mobile GSM phone by means of the Wireless Application Protocol (WAP). This paper investigates the performance of WAP over the short message service. A suitable traffic model has been identified and a GSM signalling configuration has been selected to provide the information service at the city level. Accordingly it has been possible to evaluate the delay required to transmit WAP decks to users. This study is particularly interesting because it provides a tool for sizing the envisaged tourist service.

1. Introduction

At present mobile communication systems are faced with the need to provide a wireless access to the Internet. A recently arisen solution is given by the Wireless Application Protocol (WAP) [1],[2] that allows a mobile terminal with a little display to browse Web pages or suitably designed pages in the Wireless Markup Language (WML). The WAP protocol is air interface independent and allows the easy integration of multimedia services and customised applications. WAP contains the definition of a microbrowser according to which WML and WMLScript are interpreted in the handset and presented to the user. Since a mobile user can not use a QWERTY keyboard or a mouse, WML documents are structured into a set of well-defined units of user interactions called cards. A single collection of cards is called deck, which is the unit of content transmission and is identified by a Uniform Resource Locator (URL). After loading a deck, a user agent displays the first card; then, the user decides whether to proceed or not to the next card of the same deck. Some critical aspects that need to be investigated for the adoption of WAP are as follows:

? The robustness of the WAP protocol in the presence of frequent packet errors. ? Integration of multimedia services over WAP.

? The achievable Quality of Service (QoS) of applications over WAP (e.g., the prompt delivery of

the requested WML decks) depending on the bearer services used in the mobile network.

This paper focuses on the last aspect and assumes the use of a Global System for Mobile communications (GSM) network [3]. By making minimal demands to the air interface, the WAP protocol can also operate on low-bandwidth, non-IP GSM bearers such Short Message Service (SMS), or Unstructured Supplementary Service Data (USSD) channel [4],[5]. Referring to the SMS case, the QoS evaluation for the services provided through WAP is particularly interesting on the basis of the following aspects: (i) each SMS message is limited to a maximum of 140 bytes; ( ii) the logical channel used to convey SMS messages has a slow throughput; (iii) WML decks are segmented in several concatenated SMS messages; (iv) the transmission of each SMS requires the set-up of a connection between the mobile terminal and the base station. Hence, it is evident the interest for the evaluation of the mean delay to browse a WML deck with GSM-SMS.

This study is an evolution of the techniques studied in the HIPS Project (Hyper-Interaction within Physical Spaces of the ESPRIT Program) and it is related to the Information Society Technologies Project, called Personalised Access to Local Information and services for tOurists (PALIO), within the th

5 Research Framework of the European Commission. This Project is in co-operation with an Italian GSM network operator. PALIO aims at providing multimedia services to tourists for facilitating their visit of a city. At present, tourist information is static and mainly available through paper guides, Internet sites and information points located in specific positions. The PALIO system proposes a wireless access to multimedia services through the mobile phone of the tourist. By means of the WAP protocol a user will be able to receive on a common mobile phone all relevant information stored in a suitable server at the city level. The main experimental services considered by the PALIO system deal with location-dependent personalised information for tourists.

This paper represents a preliminary study related to the mobile browsing of multimedia contents, an interesting perspective for future generation cellular systems.

2. The considered scenario for WAP

We have envisaged a tourist service supported by WAP where WML pages are hosted in a suitable WAP server at the city level. The lowest layer specified by the WAP protocol stack is the Wireless Datagram Protocol (WDP) that corresponds to the transport layer of the ISO -OSI reference model. Hence, WAP can be supported by different bearer services in various network infrastructures. Let us refer to the SMS service in GSM networks; we have made such assumption to evaluate the service potentialities in a rapidly deployable configuration that can be supported by all the GSM networks. The GSM standard envisages various data formats for SMS. The SMS payload is 140 bytes (including, if present, a header of 12-13 bytes) [5],[6]. Even if the SMS channel offers a very low throughput, it must be considered that a WML deck is encoded using a compact binary representation (tokenised form) based on the WAP Binary XML (Extensible Markup Language) content format [7], thus reducing the deck size on the air interface.

The communication between the mobile user and the WAP server occurs through the Short Message Service Centre (SMSC) which is specific for the considered service. The system architecture is shown in Fig. 1, where ad hoc SMSC and local WAP server are used in order to reduce the network response time and in order to tailor the WAP server contents to the tourist service managed at the local level. In order to access the services, the user must set on its mobile phone the number of the SMSC.

The user may send its request for a given WML deck with few SMS [2]. The WAP server responses by sending the WML deck in a suitable datagram (the maximum datagram length supported by concatenated SMS is 255 SMS). On the air interface datagrams are compressed according to [7] and are fragmented in SMS sent to the mobile user (here assumed not involved in a phone call) through a Stand alone Dedicated Control Channel (SDCCH). SDCCH is a signalling channel that is organised according to a 51-multiframe structure. In particular, in this study we have considered that the first slot

186

of each frame (on only one carrier in the cell) transports the following logic GSM downlink (i.e., from the base station to the mobile users) channels: FCCH+SCH+CCCH+BCCH+SDCCH/4+SACCH/4 [3]. Moreover, we have assumed a configuration where there are only four SDCCH channels, on four consecutive frames in the 51-multiframe. According to the GSM channel coding scheme [8], we have considered that each SDCCH burst conveys 23 information bytes, so that about 6 SDCCH bursts (on a given SDCCH channel) are required to transfer a 140-byte SMS on the air interface.

In this study we have neglected the set-up times for the SDCCH assignment to convey SMS messages in order to focus only on the SDCCH traffic capacity. Hence, for each WML deck requested by the user, the WAP server sends a datagram through a given SDCCH channel as soon as available (we have neglected any other signalling messages to be transported by SDCCH). A First-Input First-Output (FIFO) queuing scheme has been assumed for the datagram transmission on SDCCH channel.

3. Traffic model for the mobile browsing with WAP

In order to evaluate the performance of WAP over GSM -SMS, we consider the following downlink traffic model for each browsing user. We assume that the user has started a browsing session and that receives WML decks through a given SDCCH channel. In what follows, we have adapted the models in [9] taking into account the considered scenario. Each user is associated with an on/off downlink traffic generator alternating between an active period (i.e., when the user requests decks) and an idle phase (that corresponds to a pause in the browsing when the user reads the received decks). The duration of an idle phase is assumed to be exponentially distributed with mean value 1/μ id = 20 s.

During an active period, the user browses a number of decks, which is assumed geometrically distributed with mean value Nd. Referring to our envisaged tourist service based on WAP over GSM-SMS, we have considered cases with 2, 3 and 4 decks per active period (i.e., a short period of browsing). Each deck has been considered as a datagram to be routed to the mobile user. The datagram interarrival time in the active period is exponentially distributed with mean value assumed equal to 1/μ a = 5 s. The activity factor for this type of source is ψ w = (Nd/μ a)/(Nd/μ a+1/μ id) and the mean datagram arrival rate is λ = μ a ψ w datagrams/s. Of course, parameters μ id, μ a and Nd mainly depend on the behaviour of the browsing user.

We have considered that the mean number of cards per deck depends on both the service type and the bearer service used for WAP. Therefore, we have characterised the distribution of the card length in bytes by measuring the length of more than 1000 cards on different WAP servers. The mean card length resulted as

2

567 bytes, whereas the mean squared value of a card resulted as 401340 bytes. The card length histogram in bytes has been fitted with two distributions: the gamma distribution and the Pareto one [9]. The Gamma distribution has the following probability density function:

According to the fitting process, we have obtained α = 4.4843 and β = 126.6486. The Pareto distribution has the following probability density function:

Assuming k = 80 bytes (which is about equal to the minimum measured card length), we have fitted the mean value (= 567 bytes) so that we have obtained γ = 1.1. The results of the fitting procedure and the card length histogram are shown in Fig. 2. From this graph, we note that the Pareto distribution does not allow a good fitting, whereas the Gamma distribution appears as a good solution.

We have considered that the number of cards per deck is geometrically distributed with mean value Nc (reasonable values for a prompt delivery of the deck through GSM-SMS are Nc = 3 or 4 cards/deck). According to the content encoding process performed before sending each WML deck on the air interface, we have assumed a mean deck length reduction of 80% (conservative estimate [7]).

4. Performance Analysis

A downlink performance analysis is presented in this Section to evaluate the mean datagram delay, T , as a function of the number of users involved in the WAP browsing per SDCCH channel, M . In this

study we consider that a datagram is divided in the equivalent number of “packets”, where each packet represents the information content sent on an SDCCH channel (i.e., one slot on four consecutive frames) per multiframe. Hence, a packet conveys 23x4 information bytes.

A datagram traffic source corresponds on downlink to each browsing user. Each source can be approximated by a 2-state Markov-Modulated Poisson Process (MMPP), if we make the

assumption that the times spent in the active period and an idle phase are exponentially distributed with the mean values presented in Section III. Hence, the modulating process of the resulting traffic source has been shown in Fig. 3; transitions occur at the end of multiframe periods. Accordingly, we have adopted the following system model [10]: ??2??MMPP/D/1 , where: ??2??MMPP stands for the aggregation of MMdownlink traffic sources each of them with a 2-state MMPP arrival process of datagrams with a length characterised as explained in the previous Section; D is the deterministic packet service time; “1” means that only one packet can be transmitted per multiframe.