Anshuman Dutt

About

I am a researcher in the Data Systems Group (opens in new tab). I received my Ph.D. from Indian Institute of Science (opens in new tab), under the supervision of Prof. Jayant Haritsa (opens in new tab).

Research focus:

I have been working on research problems related to query optimization in database systems. While modern query optimizers are known for huge successes in handling complex customer queries and producing efficient execution plans in many cases, they often produce highly suboptimal execution plans in an unpredictable manner – it is widely believed that the primary reason for sub-optimal plans is errors in cardinality estimates.

Research works:

During PhD, we developed a novel technique called “Plan Bouquets” that can execute SQL queries with small upper bounds on the plan sub-optimality, by computing (at run-time) rather than estimating error-prone cardinalities.

More recently, the focus has been on ways to enhance different aspects of query optimization using machine learning (ML) techniques. Specifically, we developed techniques that leverage ML models, with low usage and training overhead, for improving the quality of cardinality estimates, and for optimization of parameterized queries