| The database infrastructure changes with the size of the BI decision-support environment, which in turn influences the selection of the DBMS, as shown in Figure 2.5. A small departmental data mart application may reside on a local file server, but a larger BI application may need the infrastructure support of an enterprise server, and very large enterprise-wide BI solutions may need to use a mainframe. Figure 2.5. Database Infrastructures  Criteria for Selecting a DBMS The following functions are important and necessary attributes of a DBMS for handling the workload of a large BI target database or very large database (VLDB): -
Degree of parallelism in handling queries and data loads -
Intelligence in handling dimensional data models and optimizers -
Database scalability -
Internet integration -
Availability of advanced index schemes -
Replication on heterogeneous platforms -
Unattended operations A DBMS is a sophisticated piece of software and consists of a number of features that need to be evaluated. Features to look for in a DBMS for BI applications are listed below. -
Network support provided by the DBMS should be compatible with the organization's data communications standards. -
Dimensional capability in the form of seamless support for fast and easy loading and maintenance of precompiled summaries is important. -
Adequate state-of-the-art triggers and stored procedures can be used as "event alerters," which trigger an action in response to a given set of circumstances. -
Administrative support features should provide for: -
- Maintenance of consistent historical data -
- Support for archiving (e.g., dropping the oldest week's data when adding the data for a new week) -
- Controls for implementing resource limits to display a warning when a query that consumes excessive resources is about to be terminated -
- Workload tracking and tuning mechanisms -
- Careful monitoring of activity and resource utilization -
Location transparency across the network must allow the access and analysis tools to retrieve data from multiple BI target databases from a single workstation. -
Future usage explosion must be supported by: -
- Effective caching and sharing of data to minimize input/output (I/O) bottlenecks -
- Effective management of task switching while running many queries concurrently -
- Compatibility with multiple processors -
Scalability requires that the DBMS has the capability to support: -
- Advanced functions for sorting and indexing -
- Fault tolerance for uninterrupted processing -
- Uninterrupted maintenance operations, such as unload, backup, and restore -
- Checkpoints, recovery, and rapid restart of interrupted operations -
Query performance optimization should address aspects of query processing (such as JOINs, sorting, and grouping) that require intensive use of the central processing unit (CPU). -
Load process and performance must address: -
- Data obtained directly from a variety of feeds, including disk files, network feeds, mainframe channel connections, and magnetic tapes -
- Complete data loading and preparation, including format conversion, integrity enforcement, and indexing -
The security system must support unique passwords, password protection, and the authorization constraints necessary for specific persons and for specific tables of the database. The system administrator should provide restricted access to the views and virtual tables. -
The data dictionary should feed into a meta data repository, and the database objects should be linked to all data objects described in the enterprise logical data model. Selecting and reevaluating the appropriate hardware, middleware, and DBMS components of the technical infrastructure are some of the most important activities on BI projects because they ensure the continued scalability and high performance of the BI applications. |
