Foundational data architecture patterns - Making Sense of NoSQL - page 39

High-level data architecture
pattern

Row store

Figure 3.1 High-level data
architecture patterns are used to
discuss the fundamental ways data is
stored in a system. Once you select a
high-level data architecture pattern,
there are many lower-level design
patterns that a system may implement.

Data definition
language

Joins Transactions

Low-level design
patterns

Typed
columns

Fine-grained
authorization

Views

3.2 Understanding the row-store design pattern

used in RDBMSs
Now that you have a basic understanding of an architectural pattern, let's look at the
concepts and principles of the row-store pattern associated with relational database
management systems. Understanding the row-store pattern and its use of joins is
essential in helping you determine if a system can scale to a large number of proces-
sors. Unfortunately, the features that make row-store systems so flexible also limit their
ability to scale.
Almost all RDBMS s store their data in a uniform object called a row . Rows consist of
data fields that are associated with a column name and a single data type. Since rows
are added and deleted as atomic units (a unit of work that's independent of any other
transaction) using insert, update, or delete commands, the technical data architecture
pattern name is called a row store , which is more commonly known as an RDBMS or SQL
database .
We should note that not all SQL databases use the row-store pattern. Some data-
bases use columns as an atomic unit of storage. As you might expect, these systems are
called column stores . Column stores shouldn't be confused with the term column family
store , which is used in Bigtable systems. Column-store systems are used when aggregate
(counts, sums, and so on) reporting speed is a priority over insert performance.

3.2.1 How row stores work

Rows are an atomic unit of data storage in RDBMS s. The general concept of a row
store is illustrated in figure 3.2.
The rows you insert comprise your tables in an RDBMS . Tables can be related to
other tables and data relationships are also stored in tables. The following list shows
how you might use an RDBMS to solve a business problem:
 A database modeling team meets with their business users. The business data is
modeled in a logical way to understand data types, groupings, and repeating
fields. When the modeling process is complete, the team has a physical table/
column model. This process is used for new data created within the organiza-
tion as well as for data that's transmitted to the organization from an external
source.