DATABASE MEETS SIMULATION: TOOLS AND TECHNIQUES
Peter J. Haas, Christopher Jermaine · 2009
We describe some recent technology for Monte Carlo analysis within the setting of a database management system. The original motivation for this work was the desire to quantify the uncertainty in the answers to database queries when the underlying input data is uncertain due to data integration, information extraction from text, and so forth. It has since become apparent that our prototype systems can also be used to process data-intensive queries in which uncertainty arises from the use of stochastic models to extrapolate missing or hypothetical data. Our systems then permit complex analysis and query processing without the need to continually transfer the data between the database and a simulation package. Our use of the map-reduce processing framework allows robust, massively parallel simulation processing on commodity hardware. We outline some key ideas and technical challenges that arise in our novel Monte Carlo environment.