What exactly is "Compute to Control Node" step in Azure SQL DW Query execution plan? Does that mean ADW is moving the data to control node and then performing the JOIN. I understand Shuffle operation which redistributes data among the compute nodes. But I did not get in what situation does the data flow from Compute to Control node for a JOIN.
Azure SQL DW Query执行计划中的“计算控制节点”步骤究竟是什么?这是否意味着ADW正在将数据移动到控制节点,然后执行JOIN。我理解Shuffle操作,它在计算节点之间重新分配数据。但是我没有得到数据从Compute到Control节点的什么情况。
All 3 high cost operations in screenshot are associated with moving 2 Fact tables and the biggest dimension tables.
屏幕截图中的所有3个高成本操作都与移动2个事实表和最大维度表相关联。
Query_Plan Thanks
1 个解决方案
#1
1
You can have portions of a query sent to the control node in operations such as PartitionMoves. For example, this might occur when you do a GroupBy on a column that's not a distribution column and the optimizer thinks the result set is small enough to send up to the control node for final aggregations.
您可以在PartitionMoves等操作中将部分查询发送到控制节点。例如,当您对不是分发列的列执行GroupBy并且优化程序认为结果集足够小以发送到控制节点以进行最终聚合时,可能会发生这种情况。
#1
1
You can have portions of a query sent to the control node in operations such as PartitionMoves. For example, this might occur when you do a GroupBy on a column that's not a distribution column and the optimizer thinks the result set is small enough to send up to the control node for final aggregations.
您可以在PartitionMoves等操作中将部分查询发送到控制节点。例如,当您对不是分发列的列执行GroupBy并且优化程序认为结果集足够小以发送到控制节点以进行最终聚合时,可能会发生这种情况。