Introduction

Over the last decade the cellular phone industry and the World Wide Web have experienced a phenomenal growth as people around the world have embraced these technologies at a remarkable rate. Today, most major wireless service providers in the United States, Europe, and Japan offer wireless Internet services and many Internet companies provide content that has been adapted to suit the limited display, bandwidth, memory, and processing power of small devices. Another emerging trend, related to wireless Internet, has to do with how users manage the gigantic information flow that the Internet provides. Realizing that users are being overwhelmed with information, several web content providers allow users to switch their data access model from polling and navigation to notifications or alerts. Instead of periodically browsing through the web sites for potentially useful information, an increasing number of users are adopting the model where they register for information in which they are interested. These users provide a callback address usually in the form of an email address, a cell-phone number, or a pager number, depending on their perceived importance of the information. Whenever the relevant event is triggered, the content provider sends a notification to the user. Examples of some US companies that provide such notifications include Yahoo Mobile, MSN Mobile, AOL Anywhere, and InfoSpace. All of these services allow users to subscribe to alerts for stock quotes, sports scores, lottery, horoscope, calendar events etc. If alert services becomes a popular form of user interaction with the web, it will be critical for content provider and content management companies to handle these notifications efficiently. Proper management of notifications involves understanding which types of notifications are popular, which types of devices are used by subscribers for receiving notifications, the frequency of sending these notifications on a per user basis, etc. In this paper, we study notification and browse services provided by a large popular commercial web site that is designed specifically for US users who access it via their cell-phones and PDAs. Unlike most previous web studies, which have analyzed browsing services provided over wired networks, we focus primarily on a web server that delivers notification and browsing services over wireless channels. We analyze notification and browser traces to understand the system load, the type of content that is accessed, and user behavior. We believe that our study is important for content providers, wireless ISPs, and web site managers. We note here that we do not study the performance of the web server subsystem or its architectural design. Instead, we use web server logs to analyze the browse and notification patterns of wireless web users. The rest of this paper is organized as follows. In Section 2 we review previous work done in the field of web trace analysis. In Section 3, we describe the different ways in which the web site is accessed, the characteristics of the data logs, and the types of analyses we carry out. We present detailed analysis of the notification and browse logs in Sections 4 and Section 5, respectively. In Section 6, we examine the degree of correlation between the usage of browse and notification services. We conclude in Section 7.