Integrating EDMSs and DBMSs - High-Performance Web Databases: Design, Development, and Deployment - page 536

At the end of this process, once the type of database management sys-
tem to be utilized is determined, one has enough information with which to
begin a physical and logical database design. The data analysis activities
should provide enough information to enable a design which will have a
high degree of predictability in terms of data access performance and data
storage size.
Data collection and normalization within any organization begins with
the analysis of the data as it exists currently. Various methodologies em-
phasize different approaches to this analysis. Some emphasize beginning
with the analysis of source documents, while others advocate analyzing
the data as it is presented to the users. For this discussion it is irrelevant
where one starts a project, what is important is that a “functional decom-
position” process is followed in all instances. Functional decomposition at-
tempts to distill the relevant data from some source (e.g., data collection
documents or presentation documents). As recently as 1 year ago, one
could have safely assumed that the documents would have been on paper;
however, today that may not necessarily be so. For the purposes of this dis-
cussion, however, the medium is irrelevant.
Once this distillation process or functional decomposition is finished,
one proceeds with a truly data-driven approach to analysis. The next step
involves grouping the data into logical groups called entities. Using a pro-
cess called normalization, one then proceeds to remove as much data re-
dundancy as possible from these entities, sometimes producing more
entities in the process. There are many good references on data normaliza-
tion techniques, and for the purposes of this article there is no requirement
to go into any more depth than this.
Once the entities are in third normal form one generally proceeds to as-
sociate each entity with the other entities using some entity-relationship
mapping technique. Entity-relationship mapping is, in general, an attempt
to reconstitute the data back into something that is meaningful to the busi-
ness where the data originated. A thorough understanding of the business
functions and processes that use the data is crucial for creating meaningful
entity-relationship maps. During this mapping process some form of quan-
tification of the entities also occurs.
The next step in data analysis is the transaction analysis. Transaction
analysis involves listing all of the business events that could trigger access
to the information within the as yet undesigned database and mapping the
flow of the transaction through the entities as it satisfies its requirement for
information. The transaction flow is dependent on the entity relationships.
Once all the transactions are mapped in this way and the quantity of each
transaction is determined, one has a good idea of how the data should be
ordered and indexed.