Database Reference
In-Depth Information
operations are executed in parallel, but in a pipelined manner. This assumes that
the RDBMS has the capability to decompose the query into subtasks where each
subtask has all the operations of index read, data read, join, and sort. Then each
subtask executes on the data in a serial fashion. In this approach, ideally the data-
base records are processed by one step and immediately given to the next step for
processing, thus avoiding wait times. Of course, in this method, the DBMS must
possess a very high level of sophistication in decomposing tasks.
Hybrid method.
In this method, the query decomposer partitions the query both
horizontally and vertically. Naturally, this approach produces the best results. You
will realize the greatest utilization of resources, optimal performance, and high
scalability.
Selection of the DBMS
Our discussions of the server hardware and the DBMS
parallel processing options must have convinced you that selection of the DBMS is
most crucial. You must choose the server hardware with the appropriate parallel
architecture. You choice of the DBMS must match with the selected server hard-
ware. These are critical decisions for your data warehouse.
Broadly, the following elements of business requirements affect the choice of the
DBMS:
Level of user experience
If the users are totally inexperienced with database systems, the DBMS must
have features to monitor and control runaway queries. On the other hand,
if many of your users were power users, then they would be formulating their
own queries. In this case, the DBMS must support easy SQL-type language
interface.
Types of queries
The DBMS must have a powerful optimizer if most of the queries are complex
and produce large result sets. Alternatively, if there is an even mix of simple
and complex queries, there must be some sort of query management in the
database software to balance the query execution.
Need for openness
The degree of openness depends on the back-end and front-end architectural
components, and those, in turn, depend on the business requirements.
Data loads
The data volumes and load frequencies determine the strengths in the areas of
data loading, recovery, and restart.
Metadata management
If your metadata component need not be elaborate, then a DBMS with an active
data dictionary may be sufficient. Let your requirements definition reflect the
type and extent of the metadata framework.
Search WWH ::
Custom Search