Log-based Middleware for Pervasive Application Sharing
Principal Investigator: Prasun Dewan
Funding Agency: The National Science Foundation
Agency Number: SCI-0229998
Abstract
Pervasive application-sharing allows users to collaborate regardless of their computing device, network, application, and location. More specifically, it allows distributed users sharing an application to (i) use stationary and mobile computers with varying capabilities such as servers, PCs, hand-helds, palmtops, and cell phones; (ii) dynamically switch computers based on their location, migrating session state from one computer to another; (iii) efficiently switch between offline and online modes; (iv) use a minimal network connecting only the computers of the currently active collaborators, and (v) change the sharing architecture to adapt to changes in network connectivity. Existing application-sharing infrastructures have not been designed to support all of these properties.
The hypothesis is "it is possible to support all of these requirements by developing a novel log-based mechanism for distributing I/O events across the network". More specifically, one can support these requirements by (I) designing a new form of logging, called object-based logging, that combines benefits of traditional logging and state transfer, thereby improving the space and time overheads of traditional logging; (2) developing a sharing mechanism based on object-based logging that can support (a) all application-sharing architectures that have been proposed or developed so far, and (b) arbitrary dynamic transitions among these architectures; (3) providing this mechanism as middleware (that is, a layer between application and system components); and (4) tying the mechanism, not to a particular system component such as a network windowing system, but instead a general design pattern for creating applications.
In this project, an attempt to verify this hypothesis by developing these four ideas, implementing them in a Java-based mechanism, and evaluating the mechanism to determine how space and time efficient it is, how much automation it provides to application developers, and how many different kinds of new and existing applications, system layers, and network configurations it can support. To perform these evaluations, the work will use this mechanism to support sharing of I/O events defined by several influential user-interface tools/frameworks including the VNC network frame buffer, the X network window system, the Java Swing toolkit, and the Model View Controller and Java Beans frameworks.
If successful, the work will lead to collaborative support that can be deployed in a much wider range of (device, network, system layer, and application) configurations and has been tested for usability by real users collaborative research and information.