Skip to content
Virtual vs Materialized Propagation in Ontology Modeling: A Practical Selection Guide
Big Data 9 min read

Virtual vs Materialized Propagation in Ontology Modeling: A Practical Selection Guide

This article explains the differences between virtual and materialized propagation in ontology modeling, detailing their storage mechanisms, real-time vs cached query performance, read-only nature, and providing a comparison table plus guidance for large enterprises versus SMEs on selecting appropriate data governance approaches.

AI Large-Model Wave and Transformation Guide
AI Large-Model Wave and Transformation Guide
AI Large-Model Wave and Transformation Guide

Understanding Shared Fields vs Propagation

Many practitioners confuse shared fields with propagated fields. A shared field is a unified field template (e.g., "security level", "owner") that multiple entity types can reference. Sharing only the field definition—type, enum, description, masking rules—each entity stores its own values independently. Shared fields do not automatically copy values from entity A to entity B.

Shared fields solve terminology consistency: ensuring the same business concept uses the same name and rules across different entities.

Propagation, by contrast, is the ability to fetch fields from associated entities along relationship links at query time. It comes in two flavors: virtual propagation and materialized propagation.

Virtual Propagation: Real-Time View, No Storage

In one sentence: real-time pass-through, read-only, no persistence.

Example: A work order links to an asset that carries a shared "security level" field. With virtual propagation enabled, opening the work order page shows the asset's security level. However:

The value is not stored in the work order table.

It is not a local field of the work order.

At query time, the system traverses the association link to read the asset's current value.

The propagated value is read-only ; modifications must be made on the source asset.

When the asset's security level changes, the next work order view reflects the update immediately.

Key characteristics:

No persistent storage, no data copies.

Real-time queries always return the latest data.

Read-only view; write access remains only on the source object.

Permissions follow the source object: lacking view permission on the asset's security level hides it even if propagation is enabled.

Deep link chains or high query volumes can introduce performance overhead.

Analogy: Screen mirroring. The TV displays the phone's screen in real time; the TV stores no video, and you cannot edit the phone's files from the TV.

Materialized Propagation: Cached Snapshots for Speed

In one sentence: pre-computed snapshots, persisted caches, used for query acceleration.

Same work order–asset scenario. With materialized propagation, the system periodically or incrementally copies the asset's security level into a derived snapshot. Work order queries read this snapshot directly.

Critical reminder: This copied data is still not a local shared field of the work order . It is a read-only derived cache. Source updates only appear after the next refresh job runs, introducing latency.

Key characteristics:

Data is persisted, creating a replica.

Fast query performance, suitable for reports, dashboards, and bulk analysis.

Inherent synchronization delay; not real-time.

Snapshot is read-only; the source remains the single source of truth.

Consumes storage and requires maintenance of refresh jobs, retry handling, and lineage tracking.

Analogy: Taking a screenshot. You capture the phone screen as an image on your computer. It opens instantly, but when the phone content changes, the screenshot stays stale until you take a new one.

Side-by-Side Comparison

Storage: Virtual – no persistence, real-time fetch; Materialized – persisted snapshot copies.

Timeliness: Virtual – real-time, source change immediately visible; Materialized – delayed, depends on refresh jobs.

Query Performance: Virtual – slower with deep link chains; Materialized – fast reads.

Editability: Virtual – read-only, must edit source; Materialized – snapshot read-only.

Typical Use Cases: Virtual – detail pages, frequent lookups, high real-time requirements; Materialized – reports, batch statistics, large-volume queries.

Important pitfall: Neither virtual nor materialized propagation automatically writes values into the target entity's shared fields. If you need upstream changes to sync into the work order's own security level field, that is a separate rule-based assignment action, not propagation.

Enterprise Modeling Choices: Large vs. Small Organizations

Large Enterprises, Classified Data, Strong Audit Requirements

Retain the full capability stack: manage local fields, shared fields, and propagated fields separately. Sensitive fields default to propagation-disabled; enabling propagation on a link requires approval. Distinguish virtual vs. materialized and apply each where appropriate. The complexity buys clear data ownership, lineage, and permission isolation, preventing uncontrolled spread of sensitive data.

Small-to-Medium Enterprises, Few Entities, Simple Link Chains

Adopt a lightweight modeling approach: prioritize shared fields, de-emphasize the standalone propagation concept.

Unify business semantics; register all common fields as shared fields first.

Encapsulate association lookups as read-only association views (the underlying capability remains, but the "propagated field" concept is not exposed separately).

Tag sensitive fields with a simple "no association carry-over" marker.

When persisted synchronization is needed, explicitly configure a sync pipeline to write into the shared field—avoid implicit propagation rules.

In short: SMEs should lower the modeling barrier first; don't impose a full complex governance stack just for architecture's sake.

Key Takeaways

Shared fields: unify field templates, govern terminology.

Virtual propagation: like screen mirroring—real-time view, no copies.

Materialized propagation: like a snapshot—cached copy, fast reads, with latency.

Propagation is fundamentally a read capability ; data synchronization writes are an independent write capability . Keep them separate.

Ontology modeling is never about maximum complexity. The right model always matches your organization's data scale, security needs, and team capabilities.

Original Source

Signed-in readers can open the original source through BestHub's protected redirect.

Sign in to view source
Republication Notice

This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactadmin@besthub.devand we will review it promptly.

Data Modelingdata platformdata governanceenterprise architectureontology modelingmaterialized propagationshared fieldsvirtual propagation
Reader feedback

How this landed with the community

Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.