Results 1 - 1 of 1
- In VLDB’ 06 , 2006
"... Measures are numeric summaries of a collection of data records produced by applying aggregation functions. Summarizing a collection of subsets of a large dataset, by computing a measure for each subset in the (typically, user-specified) collection is a fundamental problem. The multidimensional data ..."
Abstract - Cited by 2 (1 self) - Add to MetaCart
Measures are numeric summaries of a collection of data records produced by applying aggregation functions. Summarizing a collection of subsets of a large dataset, by computing a measure for each subset in the (typically, user-specified) collection is a fundamental problem. The multidimensional data model, which treats records as points in a space defined by dimension attributes, offers a natural space of data subsets to be considered as summarization candidates, and traditional SQL and OLAP constructs, such as GROUP BY and CUBE, allow us to compute measures for subsets drawn from this space. However, GROUP BY only allows us to summarize a limited collection of subsets, and CUBE summarizes all subsets in this space. Further, they restrict the measure used to summarize a data subset to be a one-step aggregation, using functions such as SUM, of field-values in the data records. In this paper, we introduce composite subset measures, computed by aggregating not only data records but also the measures of other related subsets. We allow summarization of naturally related regions in the multidimensional space, offering more flexibility than either GROUP BY or CUBE in the choice of what data subsets to summarize. Thus, our framework allows more meaningful summaries to be computed for a targeted collection of data subsets. We propose an algebra called AW-RA and an equivalent pictorial language called aggregation workflows. Aggregation workflows allow for intuitive expression of composite measure queries, and the underlying algebra is designed to facilitate efficient multi-scan execution. We describe an evaluation framework based on multiple passes of sorting and scanning over the original dataset. In each pass, several measures are evaluated simultaneously, and dependencies between these measures and containment relationships between the underlying subsets of data are orchestrated to reduce the memory footprint of the computation. We present a performance evaluation that demonstrates the benefits of our approach. 1.