The goal of the Cornell Jaguar project is to exploit the features of Java in a database system. Java's portability and security are ideal for extensible database systems. However, several systems issues arise when a database server is integrated with a Java virtual machine. These issues are explored by extending the Predator database system with Java. The mode of research involves implementation and performance measurement. The topics explored include security for user-defined Java extensions, resource management (using the techniques developed in the Cornell J-Kernel project), scalability for highly concurrent workloads, and optimization of queries with Java extensions in a client-server environment. The results from this stage of the project provide a comprehensive understanding of how to efficiently support Java extensions in a database system. Can Java be used for traditional database server functionality? The next stage of the project addresses this question. The goal is to run queries partially at the server and partially at the client, thereby optimizing execution across slow networks. Java is an ideal choice to represent portable algorithms, and the project studies different server designs for portable query processing. However, the use of Java could reduce the overall server performance. An important result of this project is a qualitative and quantitative study of the tradeoffs between the different design choices. Research results will be disseminated through publications, and by instantiation in a full-fledged database system, Predator. The new technology developed has the potential to significantly influence the next-generation of database systems.
www.cs.cornell.edu/Info/Projects/PREDATOR ; www.cs.cornell.edu/jkernel/
