Is a metadatabase just another name for metadata repository?

In broad informal usage they can overlap. The Rensselaer approach has a more specific integration architecture and knowledge-representation scope.

Does it automatically understand every local schema?

No. Models and mappings need to be created or derived and checked; ambiguous business meanings do not disappear through storage.

Does a global query guarantee one synchronized snapshot of all sources?

Not merely because metadata supports the query. Snapshot and transaction guarantees depend on the execution architecture and participating systems.

AI Glossary / Data engineering and architecture / Intermediate

Metadatabase

Definition

A metadatabase is a database organized around metadata: descriptions of other data, models and relationships. In the Metadatabase approach associated with Rensselaer Polytechnic Institute, it also represents contextual knowledge and rules to support integration across heterogeneous systems. This entry emphasizes that architectural meaning, while recognizing that the word is sometimes used more broadly for a database of metadata.

What it means

Connecting two databases is not enough to make their information compatible. Their tables can use different identifiers, structures and meanings. One system's item might represent a product design, while another's item represents an individual physical unit.

The Metadatabase approach makes descriptions of those systems and their relationships explicit, so they can support shared information use. It treats relevant models and contextual knowledge as managed information rather than leaving every integration assumption buried in application code.

That ambition is broader than maintaining a list of available tables. It does not imply that an arbitrary collection of descriptions can automatically reconcile incompatible meanings or guarantee consistent distributed updates.

How it works

The historical architecture uses a model for representing information resources and contextual knowledge, a storage structure for that metadata and management facilities for using it. The Global Information Resources Dictionary, or GIRD, is part of the original approach.

Descriptions can connect local data structures to enterprise concepts and represent rules about their relationships. A global query facility can use this information when formulating or processing requests involving multiple local systems. Active integration requires additional mechanisms that use the represented knowledge to coordinate interactions.

The essential engineering task is to establish valid mappings and rules. A stored assertion that two fields correspond is only as sound as its definition and maintenance. Changes to local systems must be reflected in the integration model, with handling for ambiguity, unavailable sources and operations that cannot safely be translated.

Three practical examples

1. Manufacturing information: A team describes the relationship between engineering part definitions and production records. The shared model explains which identifiers correspond and which describe different levels of detail.

2. Cross-system reporting: Analysts map local status codes to an agreed reporting concept before asking a question across independent databases. They retain distinctions that cannot be meaningfully combined.

3. AI-assisted query design: A prototype assistant consults documented structures and mappings when proposing a query. The metadata helps constrain interpretation, but the generated query still needs validation against the intended meaning and supported source operations.

Think of it this way

Think of a coordination handbook containing the maps, terminology and interaction rules for several workshops. Each workshop can keep its own records, while the handbook explains how those records relate.

The analogy is limited because a metadatabase can make descriptions machine-processable. A handbook sitting on a shelf does not execute a query, and a metadatabase still requires software that interprets its models and rules correctly. Recording a rule is not the same as enforcing it everywhere.

Worked example

Suppose database A records a product quantity as 12 boxes, with ten units per box. Database B records 35 individual units of the same product. A report asks for the total number of individual units.

A useful integration description identifies the product correspondence, the quantity units and the conversion rule. For this simplified case, A contributes 12 × 10 = 120 units, so the combined total is 120 + 35 = 155 units.

Adding the stored numeric values directly would produce 47, an answer that ignores their meaning. Mapping the column names alone would not prevent this error.

Now suppose A also contains a different product with eight units per box. The conversion must be associated with the correct product and applicable time or version. A single global rule saying every box contains ten units would create a plausible but wrong result. This exercise illustrates metadata requirements, not historical query syntax.

Illustrative case study

Fictional case: Meta Bridge wants to compare production data from two factories without replacing their local applications. The team first creates a shared list of column names and assumes the integration problem is solved.

During a review, staff discover that one factory's completed count includes reworked items while the other excludes them. The team records the definitions separately and develops an explicit reporting rule instead of presenting the values as directly equivalent.

It also establishes responsibility for updating the mappings when either application changes. The case illustrates the reasoning behind metadata-based integration. It illustrates ongoing integration work, not an actual deployment.

Common misconceptions and mistakes

A metadatabase is not necessarily a warehouse containing copies of every source record. The defining emphasis is on descriptions and knowledge about the sources and their relationships.

It is not synonymous with every federated query engine. Federation can use other architectures and metadata models.

Nor does a global view eliminate local autonomy or operational constraints by definition. The integration mechanism must respect the actual capabilities and rules of participating systems.

Finally, the historical capitalized Metadatabase approach should not be confused with a universally standardized product category. Always establish which meaning a document or vendor intends.

Business use

The concept is useful when planning integration across systems that must retain different structures or ownership. Begin by identifying the business question and the semantic differences that could change its answer.

Document mappings, units, identifiers, rules and exceptions, then test them against representative cases. Assign owners and version the descriptions. Maintaining integration knowledge explicitly makes assumptions easier to challenge.

Check your understanding

Two sources both have a quantity column. Is that enough information to add their values safely?

Answer: No. Check what each value counts, its unit, entity identity and relevant time or version. The metadata must express meaningful equivalence, not just matching names.

Related terms

Metadata repository
Information integration
Semantic interoperability
Knowledge representation
schema mapping
federated database
global schema
data dictionary

Learn next

Learn schema mapping next, including units and identifier correspondence. Then study semantic interoperability to distinguish structural compatibility from shared meaning. Finally, compare federation and warehousing, paying attention to where data resides and what consistency each design provides.

Semantic interoperability

Sources

Cheng Hsu, The Metadatabase Project at Rensselaer, SIGMOD Record, 1991: https://sigmodrecord.org/publications/sigmodRecord/9112/pdfs/141356.141394.pdf — Original research account read for the architecture, contextual knowledge, GIRD and global-query purpose. This article explains the historical concept without claiming current product capabilities. The numerical exercise, evaluation advice and fictional case are original.

Previous: Metadata structure definition Next: Metadirectory