We propose to develop a revolutionary planetary-scale Information Utility, enabling ad-vanced technologies to be exploited by broad communities of users, making them more effective in their daily tasks, such as problem solving and learning. Like the electric power grid, the Information Utility is everywhere, always there, enables virtually all of the tools of daily life, and is invisible to its users. It flows its resources from sources to consumers, adapting to instantaneous demand by borrowing resources across an international pool.
Unlike the electric power grid, the Information Utility we propose to build will have the ability to learn how it is used and will adapt its functions and interfaces to the demands of its users. It will achieve this through the provision of "fluid" technologies for plug-and-play component interoperation, dynamic adaptation, component self-organization and personalization, across a diversity of interconnected information devices from MEMS sensors to massive clustered servers. Information by-products of human activity will be automatically captured, stored, arranged for rapid access, and made available to others. "Always available" functionality insures that the utility exhibits graceful degradation and expansion.
To validate the architecture, we will stress it under demanding applications for rapid de-cision making and learning. In addition, we will develop new methodologies for the con-struction and administration of systems of this unprecedented scale and complexity. Our success will be measured by how effectively our architecture amplifies and leverages human intellect.
Our expedition is organized into a base program with eight optional tasks. The Base pur-sues a broad, but necessarily shallow, investigation into the architecture of the Informa-tion Utility. It incorporates existing Information Devices rather than developing new ones with innovative features that make them better suited for full-fledged integration with the Information Utility. It also limits the applications it undertakes to develop, to a bare minimum activity needed to drive the development of the underlying Utility functionality. Nevertheless, the Base provides a coherent context for developing early prototypes of many of the most critical technologies proposed in greater depth, diversity, and generality in the options.
In particular, the Base Program will focus on support for "always available" operation, the underlying enabling technology for fluid software (partitioning and management of state, data and processing placement and movement, component discovery and negotiation), and the flexible capture, self-organization, and re-use of information.
Specifically, the Base Program will develop and deploy:
Each option deepens the expedition in several dimensions. Option 1, System Architecture for Vastly Diverse Computing Devices, goes beyond traditional computing devices to de-velop radical new operating systems technology for embedded devices such as MEMS sensors. It is essential for all other options. Option 2, Oceanic Data Information Utility, is a horizontal slice through the Information Utility that develops the distributed, persistent storage system upon which the whole architecture depends. It builds on Option 1. Option 3, Information Capture and Reuse, unifies sensor technologies with the software needed to fully automate the extraction, management, and analysis of the streams of information such devices will generate. It is constructed on Options 1 and 2. Option 4, Negotiation Architecture for Cooperating Components, develops adaptive technologies to manage the negotiation and communications among components that advertise their capabilities and discover potential collaborators, communicate with them, and negotiate the means for their cooperation. Option 5, Tacit Knowledge Infrastructure, pursues a vertical cut through the architecture from applications to devices to develop the necessary compo-nents to support high speed decision making. It depends on Options 1, 2, 3, and 4. Option 6, Classroom Testbed, develops an alternative vertical slice, motivated by the design of a "smart" (sensor-, camera- and display-rich) classroom space, supporting for cooperative learning among local and remote participants. It is built on Options 1, 2, 3, 4, and 5. Op-tion 7, Scalable Heterogeneous Component-Based Design, creates the methodologies to enable the design of adaptive systems and components that span hardware and software. It provides the essential design and evaluation tools for all options. Finally, Option 8, Scaled-Up Field Trials, supports the development of larger-scale testbeds and insertion opportunities to evaluate the architecture's ability to scale-up.
Randy H. Katz, 17 July 1999, randy@cs.Berkeley.edu