Database Reference
In-Depth Information
Understanding MINDOP
448 represents one thread for every distribution on a 56-compute node
appliance. There are 8 distributions on each node. In this configuration
PDW would issue 448 separate queries in parallel against a distributed
table. Hence, MINDOP(448). Fear not, we will discuss distributions
and distributed tables shortly. The key takeaway here is that PDW
delivers an additional level of parallelism that simply isn't available in
SQL Server.
Data Warehouse Features Come First to PDW
Just think about this list of features; all were first in PDW:
• Updateable column store
• Enhanced batch mode for query processing
• Native integration with Hadoop via Polybase
• New Cardinality Estimator
• Cost-based Distributed SQL Query Engine
• New windowing functions including lag and lead
It's a pretty impressive list. I am sure that there are other examples, as well.
In the case of Polybase, it is an exclusive feature, as well, which helps to
define PDW's unique selling point (USP).
PDW and Hadoop
We just touched on this, and it is worth restating. PDW offers native
integrationwithHadoopviaPolybase—somethingthatwearegoingtoreally
dive into later in this chapter. The key takeaway here is that Polybase is
integrated into a component of PDW, not a component of SQL Server. That
component is called the Data Movement Service (DMS).
