
I will describe AQSim, an ongoing effort to design and implement a system to manage terabytes of scientific simulation data. The goal of this project is to reduce data storage requirements and access times while permitting ad-hoc queries using statistical and mathematical models of the data. Querying tera-scale data requires addressing several research challenges including the size of the data, multiple data formats, and supporting complex spatio-temporal queries. We are pursuing a multi-pronged approach to these issues. First, we create a hierarchical partitioning of the data, and model each partition, to create a multi resolution view. Currently we generate a statistical model and a wavelet model. Since obtaining a highly accurate response can require significant time, we provide the capability to trade accuracy for response time. Second, we use metadata associated with these models to facilitate processing the ad-hoc queries. This metadata helps match the user query to the appropriate model, allowing us to generate the most accurate answer within a user-specified error tolerance. We use the term "approximate" for the ad-hoc queries because of the described constraints.
Ghaleb Abdulla is a senior computer scientist in the Data Science group at LLNL. His research interests include scientific data management, database technology, text and data mining, information storage and retrieval, and digital libraries. Dr. Abdulla was very active in the WWW community developing efficient cashing algorithms for proxy servers. He participated in the W3C workload characterization group and co-authored two highly referenced papers in this field.
Dr. Abdulla earned his Ph.D. in computer science from Virginia Tech in 1998, and he obtained his M.S. in computer science from the same institution in 1993. He earned his Bachelor Degree in Electrical Engineering from Yarmouk University in Jordan. Before joining LLNL, Dr. Abdulla worked for the Dow Chemical Company as an Information Technology Specialist. He is a member of IEEE, ACM, Sigma Xi Scientific Research Society, and Upsilon Pi Upsilon Honor Society of the Computer Sciences.