NoSQL concepts - Making Sense of NoSQL - page 19

RDBMS NoSQL

User interface User interface

Services

Middle
tier

Key-value store

Object-relational Relational-object

Caching Search

Lucene

Transactions Batch

Search Blobs

MapReduce

Database
tier

Transactions Batch

Caching Triggers Documents

Figure 2.3 This figure compares application layers in RDBMSs and NoSQL systems. RDBMSs, on
the left, have focused on putting many functions into the database tier where you can guarantee
security and transactional integrity. The middle tier is used to convert objects to and from tables.
NoSQL systems, on the right, don't use object-relational mapping concepts; they move database
functions into the middle tier and leverage external services.

Remember, if you focus on a single tier you'll never get a fair comparison. When per-
forming a trade-off analysis, you should compare RDBMS s with NoSQL systems as well
as think about how repartitioning will impact functionality at each tier. This process is
complex and requires an understanding of both RDBMS and NoSQL architectures.
The following is a list of sample pros and cons you'll want to consider when perform-
ing a trade-off analysis.
RDBMS pros:
 ACID transactions at the database level makes development easier.
 Fine-grained security on columns and rows using views prevents views and
changes by unauthorized users.
 Most SQL code is portable to other SQL databases, including open source
options.
 Typed columns and constraints will validate data before it's added to the data-
base and increase data quality.
 Existing staff members are already familiar with entity-relational design and
SQL .

RDBMS cons:
 The object-relational mapping layer can be complex.
 Entity-relationship modeling must be completed before testing begins, which
slows development.
 RDBMS s don't scale out when joins are required.
 Sharding over many servers can be done but requires application code and will
be operationally inefficient.
 Full-text search requires third-party tools.
 It can be difficult to store high-variability data in tables.