Database Reference
In-Depth Information
NOTE
The fact that the DMS was chosen as the integration point means that
this feature isn't coming to the box product known as SQL Server
anytime soon. The DMS is PDW specific, so if you want to get on the
Polybase bus, you will need to buy a PDW ticket.
PDW Release Cadence
As mentioned previously a few times now, PDW is an appliance, an
integrated blend of both hardware and software. This fact has other
implications that are less obvious. For example, PDW doesn't
do
service
packs. PDW has appliance updates (AUs), with the team committed to
delivering an AU every 6 months.
AUsdifferfromservicepacksinonenotableway.Thereleasesarefunctional
and do not just contain fixes to defects. In actual fact, they are packed full
of features targeting the data warehouse work stream. Therefore, the team
can rev the product in a much more agile fashion and respond to customer
demand as appropriate.
How PDW Works
We are going to start by keeping things simple and build from there. This
section is a conceptual introduction to PDW.
PDW uses a single master node known as the
control node
as the front
door to the appliance and is the gateway to the data. The data is held on a
number of
compute nodes
. The vast majority of data processing occurs in
thecomputenodes.TheabilitytocontinuallyaddcomputenodesgivesPDW
its scale-out ability. The challenge when scaling out is that different subsets
of data are held on different nodes and may need to seamlessly move the
data to resolve a users' query. PDW achieves this with an external process
called the Data Movement Service (DMS).
