Cost models for view materialization in the cloud
Proceedings of the 2012 Joint EDBT/ICDT Workshops
| Hi-index | 0.00 |
After many years of studies, efficient data cube computation remains an open field of research due to ever-growing amounts of data. One of the most efficient algorithms (quotient cubes) is based on the notion of cube cells closure, condensing groups of cells into equivalence classes, which allows to losslessly decrease amount of data to be stored. Recently developed parallel computation framework Map-Reduce lead to a new wave of interest to large-scale algorithms for data analysis (and to so called cloud-computing paradigm). This paper is devoted to applying such approaches to data and computation intensive task of OLAP-cube computation. We show that there are two scales of Map-Reduce applicability (for local multicore or multiprocessor server and multi-server clusters), present cube construction and query processing algorithms used at the both levels. Experimental results demonstrate that algorithms are scalable.