Discussion: Column Stores vs. Row Stores
Nicholas Taylor, University of Pennsylvania

I will lead a discussion on the trade-offs between row-based storage
and column-based storage in database systems. Databases have
traditionally used row-based storage, where an entire tuple is stored
in one sequential block on disk. In column-based storage, tuples are
partitioned by attribute and then stored, so all of the values for a
given field end up together. This has advantages for certain classes
of applications, and has been a topic of a number of recent papers.

The following two papers advocate the column-based approach. The first
paper is a more general overview paper and is quite accessible. The
second is more technical and focuses on the CPU efficiency of the two
approaches. I will briefly present the key results from both before
opening the floor for a discussion (and rebuttal).

http://cs-www.cs.yale.edu/homes/dna/papers/abadi-sigmod08.pdf
http://www.cwi.nl/themes/ins1/publications/docs/ZuNeBo:DAMON:08.pdf