About a paper titiled "Memory-Efficient GroupBy-Aggregate using Compressed Buffer Trees"

[Lijie Xu]

Hi, all. I noticed a paper called "Memory-Efficient GroupBy-Aggregate using Compressed Buffer Trees" in SoCC 2013 accepted list yesterday. Check out http://www.socc2013.org/papers.

The detailed paper can be found here (http://www.cercs.gatech.edu/tech-reports/tr2012/git-cercs-12-08.pdf) and there are simplified slides (http://www.cercs.gatech.edu/iab13/material/amur.pdf).

It argues another GroupBy-Aggregate approach which is different with MapReduce's Sort and Spark's hash-based aggregatation.

It seems a new way but I cannot understand the pros and cons of this approach.

Any comments and discussion are welcome.