...
Combining a replicate operator, the aggregation-based RangeMap is generated based on a streaming algorithm to dynamically construct the histogram.
use dataverse tpch; let $rg := rg( for $d in dataset Lineitem return $d.l_extendedprice ) return $rg |
---|
3. Parallel Sort
In general, the parallel sort is divided into five stages, i.e., replicate, local aggregation, global aggregation, forward, sort and merge, to scale up the sort based on Hyracks.
A parallel sort template can be given as:
use dataverse tpch; for $d in dataset Lineitem order by $d.l_extendedprice /*+ psort */ return $d |
---|
3.1 Five stages of parallel sort
4. Binary In-Equal Join