Stavros Harizopoulos

about

Stavros received his Ph.D. in Computer Science from Carnegie Mellon, in 2005. Through 2007, he worked as a Post-Doctoral researcher at the DB group of MIT where he was an early contributor to the C-Store / H-store projects (commercialized by Vertica / VoltDB). Until recently he was a Senior Research Scientist in the Information Analytics Lab of Hewlett-Packard. Stavros's interests and expertise are in column-store databases, main-memory databases, query processing on new processor and storage technologies, and energy-efficient data management systems.

Research overview (2006 - today)

Energy-efficient database computing

Energy-efficiency in large data centers, when running data management and analysis tasks, will ultimately be controlled and improved at the software layer. Our SIGMOD'10 and VLDB'12 papers present the first comprehensive studies on the energy efficiency of single-node and multi-node database servers. CloudAlloc and Micro-Cellstores explore practical applications of energy-saving data system designs.

Query processing on new processor & storage technologies

The hardware landscape is rapidly changing. Flash SSDs, multi-core CPUs, GPU co-processing, deep memory hierarchies, are all here; PC RAM and memristors are around the corner. Our work in this area studies how data management platforms can best benefit by new and upcoming technologies.

Main-memory scale-out transaction processing

Main-memory scale-out SQL/noSQL engines are taking off. Our papers aim at paving the way through a design starting point, a very fast reference implementation, and a detailed performance analysis of OLTP engine components.

Column-oriented database systems (column-stores)

Columnar storage is emerging as the prevalent format for building high-performance data-analytics engines. Our VLDB’06 paper analyzed the performance tradeoffs of columns vs. rows, and our SIGMOD’09 paper explored columnar storage on SSDs along with columnar out-of-core joins.