Generalized MD-Joins: Evaluation and Reduction to SQL

Using grouping variables to express complex decision support queries

“…4 Note that although Definition 3.1 states that for each row r of B we identify a set of tuples of R, in this algorithm we follow the reverse direction (as in hashbased approaches): we scan the detail table R and loop over all tuples of B to identify matches based on condition h. If a match is detected, we update the aggregate columns appropriately. This kind of processing has several implications in performance and is appropriate for data stream processing, discussed in Section 6.…”

“…For example, if the detail relations reside in different disks, then it is possible to simultaneously scan both relations and compute the answer in one pass. Generalized MD-joins have been discussed in [4][5][6].…”

“…As a result, it is essential to be able to express grouping variables in relational algebra. In this section we present an operator, called MD-join [13,[4][5][6], which generalizes the conventional notion of group-by [21]: it clearly distinguishes between the definition of the ''base values'' used to aggregate-by and the actual computation of aggregates of these. Grouping variables represent the latter.…”

Partitioned optimization of complex queries

“…The MD-Join operator was introduced in [17] and was further explored and generalized in [36,37]. It represents a particular, generalized kind of groupwise processing.…”

“…Once again, groupwise processing can be described via a relational operator within the standard optimizer of any system. The MD-join is generalized to GMD-joins in [36] and its usage in OLAP in distributed data warehouses is assessed in [37].…”