Proposal Number: IIS-0713409 Proposal Title: "Staged Database Systems: Maximizing Locality through Service-based Data Management" PI: Anastasia Ailamaki Affiliation: Computer Science Department School of Computer Science Carnegie Mellon University
Database software architectures have traditionally followed the client-server paradigm, where requests are treated as processes and contect-switching is performed by the operating system. Unfortunately, microarchitectural design moves towards the opposite direction: Memory hierarchies are becoming deeper, requiring better software locality, and chip multiprocessors enable unprecedented degrees of parallelism, unacheavable by current database systems.
The proposal details the design, implementation, and evaluation of a Staged Database System, which uses the Service-Oriented Architectures (SOA) paradigm for high-performance relational Database Management Systems (DBMS). The new system will (a) alleviate the mismatch between DBMS software behavior and modern CPU architectural features, (b) provide a rich infrastructure for optimizing resource scheduling, and (c) improve scalability and configurability of data management on multiprocessors and networks of workstations.
The intellectual merit of the project is in that (a) it promotes interdisciplinary work across several areas (databases, compilers, computer architecture, performance modeling), and (b) that the scientific papers stemming from the results will likely influence the way people think about building large systems (not only database systems). Such a system will have direct societal impact to the millions of people who are now using database backends directly or indirectly through web applications and services. Finally, the project will have significant educational impact, as it will serve as a vehicle to (a) train graduate students to building systems and to experiment both in a real system testbed and in simulation, (b) develop new interdisciplinary courses, and (c) disseminate the research results through talks, seminars, and publications.
Finally, the PI has long demonstrated her dedication to training students from underrepresented groups. She has been offering seminar talks, organized panels, and regularly attended the Grace Hopper Conference for Women in Computing. She has also had numerous female PhD and intern students, and she intends to continue devoting resources to attracting women to computer systems research.
Further information concerning the project can be found at the web page: www.cs.cmu.edu/~StagedDB/
