When processing graphs, SQL-GR spools graphs to and from disk. In addition, SQL-GR allows vertices to opt in and out of processing.
SQL-GR initializes an in-memory representation of a graph from one or more co-partitioned input tables. SQL-GR invokes a user-defined vertex constructor once per input partition. The number of cogroup partitions after the initialization phase is equivalent to the number of vWorkers because the distribution function used to map vertices to graph partitions is the same hash function and bucket map combination used to partition and repartition Aster instance tables.