The demand for network resources is steadily increasing. End user experience with a network service is often limited by the weakest link along the path from the client to the server. With the client-server computing model still dominant in today's Internet, traffic tends to converge at the edge of the network, creating performance bottlenecks. As a consequence, performance is still a crucial part of any Internet service.
The performance issue becomes even more important in the face of the proliferation of mobile (Internet) services. Mobile devices often rely on wireless Internet connections that have much more limited bandwidth and are much less reliable. Work load characteristics and usage patterns are substantially different from those of traditional wired and desktop applications. Yet all of these have to work with technologies that were conceived in the context of wired services. This poses significant challenges to the designers, developers, and deployers of mobile services.
In this research, we begin by surveying existing mobile services, and identify and examine specific performance issues faced by these services. In particular, we study the service models, system architectures, and most importantly, the workload characterizations of existing mobile services and their implications for today's Internet architecture, applications, algorithms, and protocols. Finally, we are beginning a study of iMobile (a.k.a. AT&T Mobile Network)--a mobile service platform that allows mobile devices to communicate with each other and to securely access corporate and Internet contents and services. We are working with AT&T to collect and analyze trace data from online iMobile gateways and servers.