High-performance Main-memory Data Management

  • Date
    February 14, 2013
  • Time
    2:30 pm
  • Location
    366 WVH


Decision makers today want to analyze constantly evolving datasets of unprecedented volume and complexity in real time. This poses a significant challenge for the underlying data management system. In the past, data processing could scale to meet the growing demand with few changes to the underlying software components mainly due to a sustained improvement in single-threaded CPU performance. Because of fundamental technological limitations, however, single-processor performance has recently been increasing much more slowly than in the past.

It is not uncommon today for a single database server to be able to concurrently execute instructions from hundreds of threads and store terabytes of data in main memory. Commercial database management systems, however, have not been designed for such hardware; they treat main memory as a vast software-controlled cache, and rely on multiple concurrent requests to fully utilize a modern system. My work advocates for significantly improving data processing efficiency through a redesign of the data management software to better leverage modern hardware.

In this talk, I will discuss how to improve the performance of a database management system for memory-resident data by redesigning two core components: the storage engine that is responsible for storing and indexing data, and the query execution engine that is responsible for evaluating relational operators. I will then present performance results from two prototype data processing systems. The first is Pythia, a parallel main-memory query execution engine that is designed for modern multi-socket, multi-core hardware. The second is Hekaton, a transaction processing engine for memory-resident data, which is currently being commercialized by Microsoft in the next release of Microsoft SQL Server.

Brief Biography

Spyros Blanas is a Ph.D. candidate in the department of Computer Sciences at the University of Wisconsin-Madison, and is advised by Jignesh M. Patel. His research aims to make data processing more efficient by reconsidering how data management software interacts with modern hardware. During his Ph.D. studies, Spyros has extensively collaborated with researchers and professionals in the computer industry through several industrial internships at Microsoft and IBM.