Database Reference
In-Depth Information
WARNING
Directly Accessing PDW's SQL Servers and the Shell Database PDW
does not let you directly access the SQL Servers either on the control
node or the compute nodes. This is to ensure that no inadvertent
changes are made that could damage PDW and potentially void the
warranty. Consequently, whilst it is an important component to
understand, the shell database is not directly accessible by end users.
It's created purely for PDW to use.
Each compute node also has a database created. These databases are where
all the user data is stored. These databases have a rather interesting
configuration. They consist of 10 separate filegroups that are key to PDW's
parallelism. Each database on each compute node has all 10 filegroups. I've
detailed them in the all in the matrix shown in
Figure 10.3
.
Figure 10.3
The bucket matrix created by PDW for holding user data
I have listed out all the filegroups on the
x
-axis of the matrix and shown
compute nodes on the
y
-axis. This symbolizes something important. Every
compute node has its own database each with their own set of filegroups.
The matrix represents the total number of
buckets
available to PDW for
depositingdata.ThisreallyisthekeytohowPDWworks.Themorecompute
nodes you have, the more buckets.
The filegroups on the x-axis start with some unusually named ones DIST_A
- DIST_H. These are called
distributions
and are designed for a certain type
of table called a
distributed table
. The next filegroup is called
replicated
.
You will notice that there is only one of these. That is significant. This
filegroup holds
replicated tables
.
