2012 Poster Sessions : Efficiently Updating Column Stores

Student Name : Abhijeet Mohapatra
Advisor : Mike Genesereth
Research Areas: Artificial Intelligence
In this poster we present a technique to support fast updates and look ups of values based on their offsets in run-length encoded sequences. This technique is of substantial interest for column stores which extensively use run-length encoding and bit-map encoding to compress attribute values. While columnar databases update a relation in time which is linear in the number of tuples, our technique can significantly lower the update complexity to be logarithmic in the number of tuples and avoid the cost of re-compressing updated tuples.

Abhijeet Mohapatra is a 4th year PhD candidate in the Logic Group advised by Professor Michael Genesereth. His work focuses on data integration and view-based updates.