Proceedings of the 6th Hellenic Data Management Symposium (HDMS2007), Athens, Greece, July 2007.
Nikos Hardavellas, Ippokratis Pandis, Ryan Johnson, Naju G. Mancheril, Stavros Harizopoulos*, Anastasia Ailamaki and Babak Falsafi
Carnegie Mellon University
* MIT
http://www.pdl.cmu.edu/
Prior research shows that database system performance is
dominated by off-chip data stalls, resulting in a concerted effort to
bring data into on-chip caches. At the same time, high levels of
integration have enabled the advent of chip multiprocessors and
increasingly large (and slow) on-chip caches. These two trends
pose the imminent technical and research challenge of adapting
high-performance data management software to a shifting
hardware landscape.
In this paper we characterize the performance of a commercial
database server running on emerging chip multiprocessor
technologies. We find that the major bottleneck of current software
is data cache stalls, with L2 hit stalls rising from oblivion to
become the dominant execution time component in some
cases. We analyze the source of this shift and derive a list of
features for future database designs to attain maximum
performance. Towards this direction, we propose the adoption of
staged database system designs to achieve high performance on
chip multiprocessors. We present the basic principles of staged
databases and an initial implementation of such a system, called
Cordoba.
FULL PAPER: pdf