|
|
|
This is a milestone paper on implementation of data warehouse using MOLAP technology or more concretely on data cube computation.
It shows how array-based aggregation of multi-dimensional data warehouse can be implemented efficiently using sparse array technique, plus their new chunk-based compression technique, and appropriate ordering of memory-resident portion of dimensions in the multiway, and multi-dimensional array aggregation computation.
The paper also compares ROLAP-based computation and shows the high performance of their array-based aggregation method.
The presentation is crystal clear and I really enjoy reading the paper!
I should note that complete computation of data cube may lead to explosive size of a data cube especially when the number of dimensions grows. Therefore, there is a serious limitation on the number of dimensions a MOLAP method or any other method may handle. This problem has been addressed in Ken Ross' VLDB'97 paper [2]. Another interesting solution has been proposed at SIGMOD'99 by Kevin Beyer and Raghu Ramakrishnan [3].
I have not seen new work in this direction: some nice results on array-based computation of ICEBERG cubes. Can someone give more pointers on it?
Copyright © 1999 by the author(s). Review published with permission.