As the technology advances, and Web technologies are becoming more and more ubiquitous there is an ever-growing need to deal with vast amount of data collected. In order to efficiently collect, store and extract information from these data, Information Architecture disciplines have developed and introduce several techniques which facilitate this. One of the most widely used and supported technology is the Relational Database Management Systems or RDBMS.
RDBMS implement special data structures represented as tables and relationships between those tables and the data they contain are enforced using special indexes commonly known as primary and foreign keys. The database schema supporting a IT system is one of the core subsystems; a good DB design will help towards efficient response times in user request. Modern RDBMS are incorporating subsystems and mechanisms that help towards availability, reliability and scalability (Connolly & Begg, Addison Wesley, 2005).
In contrast with Object Oriented architectures which use custom objects and data types, RDBMS are based on user created entities as tables with attributes as columns and the data contained represented with integral data types such as integers, decimal and varchar. Such data types can be very flexible in the storage they demand (SQL Server 2008 Books Online).
From my experience, the choice of the datatypes used when designing a DBschema is very crucial. It is tempting to always opt for a bigger datatype (i.e. int instead of smallint or tinyint) even when you know that this column won’t store any values bigger than 1 or 2 bytes. This has pros and cons: doing so will ease the future scaling of the system if the requirements change, but it won’t produce any errors if the client application tries to store values that are inconsistent with the behaviour of the entity the table represents, violating the data integrity without informing the user, thus more prone to error results.
No comments:
Post a Comment