SCALABLE IMPLEMENTATIONS OF MULTI-DIMENSIONAL AGGREGATIONS WITH INPUT BLENDING IN DISTRIBUTED QUERY PROCESSING SYSTEMS

摘要

Systems and devices implement scalable implementations of multi-dimensional aggregations with input blending in distributed query processing systems. Multi-dimensional aggregations for identifiers/values designated fields in datasets are performed based on keys. Datasets are sorted by identifier/value and divided into first partitions. Each row of data with a specific sorted-by-identifier/value is only present in one of the first partitions. Keys are generated from each combination of two or more dataset fields, and a blended table of data is generated over the partitions based on each different key combination. Designated data field characteristics are determined for the blended table based on the different key combinations. The characteristics are divided into second partitions based on the keys, where each key is present in only one of the second partitions. A final designated data field characteristic is determined for each row of data in each of the second partitions as the multi-dimensional aggregation.