Database Management Systems

Metadata provides context to data by describing its meaning, structure, and relationships, acting as the semantic layer between IT systems and business users. Metadata answers questions like what the data means, where it came from, and how it relates to other data. Master data refers to core business entities, not contextual information.

When an enterprise utilizes multiple heterogeneous ETL and reporting tools, a distributed metadata management architecture is recommended to allow individual tools to maintain their metadata while enabling sharing.

Metadata Management is applicable across all layers of the data architecture - the source layer (describing source systems), the data warehouse (describing warehouse structures and mappings), and the report layer (describing report definitions and business terms). It provides context and documentation throughout the data flow, making all layers correct.

Data warehouses use dimensional modeling (star/snowflake schemas) which intentionally denormalize data for query performance. While normalization is possible in theory, it's rarely used in practice because it would slow down the complex analytical queries that warehouses are designed for.

The Operational Data Store (ODS) is specifically designed to hold current, operational data at the most detailed level of granularity for near real-time decision support. Unlike the data warehouse (which stores historical data), staging area (temporary), or data mart (subset), the ODS maintains the lowest-level current DSS data.

In a star schema, each dimension table has a one-to-many relationship with the central fact table. A single dimension record (like one product) relates to many fact records (many sales transactions). One-to-many is the defining characteristic that enables efficient querying.

Data Quality Management is the systematic process of identifying, correcting, and preventing data defects to ensure data is fit for purpose. Unlike transformations (format changes), mining (pattern discovery), or security (access control), DQM directly targets bad data elimination.

Business Intelligence systems integrate data from multiple sources and transform it into meaningful information for decision-making. Option B is incorrect because BI analyzes both historical AND current/predictive data. Option C is wrong because BI applies business rules rather than defining them.

In BI architecture, data starts dispersed across source systems, gets collected into the staging/warehouse, gets integrated into a consistent format, becomes information through processing, and finally becomes knowledge through analysis. This progression matches option A exactly.

Data Integration is fundamentally about combining data from multiple disparate sources (databases, files, APIs) into a unified, consistent view in a central repository like a data warehouse. Options B and C describe aggregation/drill-down, not integration. Option D describes the opposite of integration.

MOLAP (Multidimensional OLAP) pre-aggregates data into cubes for fast query performance, making it ideal for small datasets with complex analytical models. ROLAP is better for large data, and HOLAP combines both but adds complexity. For small data + complex models, MOLAP's pre-computation advantage shines.

The Operational Data Store (ODS) serves three key purposes: enabling detailed operational decision-making with current data, supporting operational reporting for day-to-day monitoring, and providing drill-through capability from summary to detail levels. All options are valid ODS functions.

ELT (Extract, Load, Transform) is recommended for small to medium data volumes with good quality and off-peak loads because it loads data first into the target system, then leverages the target system's processing power for transformation. This is efficient when the target system (like a modern cloud data warehouse) has strong transformation capabilities and the load window is flexible.

The Registry architectural style in Master Data Management (MDM) provides a real-time central reference by maintaining an index or pointer to master data stored in source systems, without physically consolidating it. This enables real-time access to current data. Consolidation physically stores copies, Coexistence maintains separate golden records, and Transaction style focuses on operational updates.