First, future information appliances will be increasingly diversified. While a short time ago most Internet clients were general purpose PCs and workstations, in the near future, cell-phones, PDAs, transportation vehicles, home appliances, smart-buildings, and traffic sensors will all be connected to the global information utility. Many of these devices might not need full-fledged general-purpose processors. They might use different protocols to connect to a network gateway, and may access a different type of content (other than HTML, WML and similar formats designed for human consumption). Furthermore, low power devices such as web-enabled cell phones will impose different performance requirements on the infrastructure of the information utility than high-end workstations. This heterogeneity must be accommodated by the utility via a form of application-level differentiated services access. A one-size-fits-all information retrieval service (such as that provided by the current web) will no longer be adequate.
The second challenge that the information utility must meet is to deal efficiently with very large amounts of content that are updated at high rates. The current web infrastructure and web caching protocols are optimized for static content. In contrast, most of tomorrow's information content will be real-time such as current stock quotes, current readings of traffic sensors, and current state of various appliances. This trend has already manifested itself in the increase of dynamically-generated web content today. An efficient infrastructure is needed for propagating raw data updates from distributed sources and sensors to the client population on need to know basis.
We address the aforementioned challenges by focusing on application-layer components; namely databases, web servers and information proxies. The project has two expected contributions. First, we argue for differentiated web services as a key element of the Internet information utility to deal with the appliance heterogeneity problem. In this framework, servers and proxy caches can give preferential treatment to specified types of content. We shall design, implement, and evaluate mechanisms for performance differentiation in the server and web proxy cache. Second, we develop an integrated framework for consistency management in which databases that contain volatile information communicate using appropriate consistency management protocols with front-end servers and content caches to maximize the efficiency of dissemination of volatile information while maintaining a specified degree of consistency. In this framework, the same piece of information may be accessed by a workstation, a cell-phone or a home appliance. The content is selected automatically at the database query level depending on the client's resource limitations and consistency requirements. We shall evaluate the impact of differentiated web services and dynamic content dissemination on the performance of information access and generate prototypical web servers and proxy caches that implement our performance-enhancing extensions. We expect our prototypes to improve significantly information access performance.