The very point of database management systems is to shield an application developer from the complexities of efficient data management. Convinced that "one size does not fit all”, however, most DBMSs make assumptions about factors such as data model, usage scenario or targeted hardware. In particular when managing memory-resident data, flexilibity is usually sacrificed for common-case performance. However, recent techniques such as adaptive data management and just-in-time query compilation have changed the rules of the game. These techniques make it possible to build DBMSs that make much fewer assumptions about their environment and, thus, are more resilient to changes in the setup. In my talk, I will discuss three major techniques that were enabled by just-in-time query compilation: partially decomposed storage, efficient CPU/GPU co-processing and efficient data-model agnostic query processing. While the former two are published work, the latter is part of an ongoing effort at MIT.

Holger Pirk is a Postdoc in the Database Group at MIT CSAIL. He just finished writing his PhD thesis in the Database Architectures group at CWI in Amsterdam and will defend it on May 1st, 2015 in Amsterdam. He received his master’s degree (Diplom) in computer science at Humboldt-Universität zu Berlin in 2010. His research interests lie in analytical query processing on memory-resident data. In particular, he studies storage schemes and processing models for modern hardware.