ASSET queries: a declarative alternative to MapReduce

Today's complex world requires state-of-the-art data analysis over truly massive data sets. These data sets can be stored persistently in databases or flat files, or can be generated in realtime in a continuous manner. An associated set is a collection of data sets, annotated by the values of a domain D. These data sets are populated using a data source according to a condition θ and the annotated value. An ASsociated SET (ASSET) query consists of repeated, successive, interrelated definitions of associated sets, put together in a column-wise fashion, resembling a spreadsheet document. We present DataMingler, a powerful GUI to express and manage ASSET queries, data sources and aggregate functions and the ASSET Query Engine (QE) to efficiently evaluate ASSET queries. We argue that ASSET queries: a) constitute a useful class of OLAP queries, b) are suitable for distributed processing settings, and c) extend the MapReduce paradigm in a declarative way.