|
Tuesday, March 8, 2005 |
|
|
|
|
Adaptive Query Processing |
|
|
Database systems enable complex queries to be expressed in a convenient high-level declarative language. Queries are translated to efficient execution plans based on statistical estimates of the underlying data and system characteristics. With ever-increasing data, query, and system sizes and complexity, statistical estimates can be error-prone, leading to poor execution plan decisions. The problem is exacerbated in systems that process queries over continuously-arriving data streams -- e.g., stock tickers in financial analysis -- because statistics and system conditions may change over time. We describe how adaptive query processing as a basic technique can address these problems effectively. We first consider data stream systems, presenting a query processing architecture that adapts execution plans quickly and efficiently when conditions change over time. We then present a new approach for conventional database systems that chooses execution plans that are more resilient to statistical errors and more flexible to run-time replanning. We conclude by outlining research directions for adaptive processing as a key component of future database systems. Bio: Shivnath Babu is a Ph.D. candidate in the Department of Computer Science at Stanford University. He is working in the area of data stream systems and self-managing database systems. He received his Bachelors degree from the Indian Institute of Technology, Madras, in 1999. |
|