Our goal is to build a database framework in which the database predicts user queries and builds indexes in advance to allow future queries to run faster.

Database indexes have the reputation of being slow to build and unwieldy to manage over time. We propose Chameleon: a database system which uses lightweight bitmap indexes and preemptively builds indexes based on predictive models of user queries. In Chameleon, the overhead cost of building these bitmap indexes is virtually nil, and the predictive models have shown to hold advantages over simple historical cache-like policies.

Research Areas


Aditya Parameswaran