As a first step in executing a parallel sort, Adaptive Server creates a distribution map. If the sort is performed as part of a merge join, and there are statistics on the join columns, the histograms are used to build the distribution map. For other sorts, Adaptive Server selects and sorts a random sample of data from the input table. This distribution information—referred to as the distribution map—is used in the second sort step to divide the input data into equally sized ranges during the next phase of the parallel sort process.
The distribution map contains a key value for the highest key that is assigned to each range, except the final range in the table. In Figure 9-1, the distribution map shows that all values less than or equal to 4 are assigned to the first range and that all values greater than 4 are assigned to the second range.
