Computer hardware systems have evolved from monolithic machines in which each component performed one specific task, to complex systems with a wide range of heterogeneous components such as spinning disks, SSDs, RAM, and CPUs. However, the improvements in hardware technology have not resulted in an more efficient interaction when these various memory components work together to accomplish a certain task.
In his thesis, Holger Pirk, PhD student at the Centrum Wiskunde & Informatica (CWI) in Amsterdam, investigates the management of relational data in such a modern computer hardware and addresses the question how data processing performance can be improved. The results of his thesis are in particular of interest for companies that rely on fast data analysis for their services such as internet and credit card companies.
In his research, Pirk relates the asymmetric characteristics of the memory components (such as fast & small vs. slow & large) with the asymmetry in data access (such as sequential vs. random) and explores different strategies to identify how asymmetries in data management applications can be matched with asymmetries in the memory landscape. As a result these strategies can used to improve the performance of hardware systems.
Another problem that he addresses in his research, is the connection between data storage and query processing. Based on the just-in–time compilation as query processing model, Pirk developed a new processing model that makes efficient use of the asymmetries in the underlying data and memory components.
The research is an important step in the development of better solutions for faster processing and analyzing larger data sets. Pirk will defend his dissertation "Waste not, want not!`- Managing Relational data in Asymmetric Memories”, 1 May, 2015, at the University of Amsterdam. Promotors are Prof.dr. Martin Kersten and Prof. Dr. Stefan Manegold.
Location: Agnietenkapel, Amsterdam
Time: 12:00 am