Big Data, Data Warehouse and Data Marts
On this page
2.2 Big Data, Data Warehouse and Data Marts
Suggested retrieval lesson: 20–30 minutes.
Recall first
- Why should an operational database not automatically be used for every analytical query?
- Distinguish a data warehouse from a data mart.
- What do the “5 Vs” of big data describe?
Commit before reading.
Three related but different ideas
A data warehouse is a centralized analytical repository that integrates historical data from multiple operational and external sources. It is organized for reporting, comparison, and analysis rather than high-volume day-to-day transaction entry. Data are commonly extracted, transformed/cleaned, and loaded (ETL) into the warehouse; modern systems may extract, load, and transform (ELT) inside the target platform. IBM describes warehouses as systems for consolidating data for analytics and reporting. IBM, “What is a data warehouse?”
A data mart is a smaller, subject- or department-oriented slice of analytical data—for example, finance, sales, or HR. A dependent mart is fed from an enterprise warehouse; an independent mart may be built directly from sources. A mart can improve focus and access speed but may create silos and inconsistent definitions if departments build separate versions. IBM, “What is a data mart?”
Big data refers to data whose scale, speed, diversity, or other properties challenge conventional tools and practices. A useful exam mnemonic is the 5 Vs: volume (amount), velocity (speed), variety (formats/sources), veracity (trust/quality), and value (useful benefit). The Vs are a framing device, not a claim that every source uses exactly five. IBM, “What is big data?”
Operational database versus warehouse/mart
| Question | Operational database (OLTP) | Warehouse/data mart (OLAP) |
|---|---|---|
| Main purpose | Run current business transactions | Analyze history and trends |
| Typical workload | Many short inserts/updates | Fewer complex reads/aggregations |
| Data | Current, detailed, normalized | Historical, integrated, often dimensional |
| User | Frontline application | Managers, analysts, BI tools |
| Risk if mixed carelessly | Analytical queries slow live service | — |
A warehouse may use a star schema: a central fact table (measures such as sales amount) linked to dimension tables (time, product, customer, location). A data lake is a related but distinct repository that often keeps large volumes of raw structured, semi-structured, and unstructured data; it is not synonymous with a curated warehouse.
Worked example: university enrollment
The student information system must process registration and fee payments accurately in real time: it is OLTP. Each night, approved data are extracted, standardized (for example, course codes and dates), checked, and loaded into a warehouse. The registrar mart contains enrollment and completion measures by course, term, program, and campus. A data-science team may also analyze clickstream and text in a lake. The warehouse answers “enrollment trend by program”; the live database answers “is this seat available now?”
Exercise — reveal after committing
A sales department creates a mart directly from its CRM and reports 12% growth. Finance’s warehouse reports 9% because returns and canceled orders are handled differently. Diagnose the problem and propose one fix.
Revealed answer: The marts use different definitions, sources, or transformation rules—an analytical data-governance problem, not merely arithmetic. Establish shared metric definitions and lineage, preferably publish the mart from a governed warehouse or reconcile both pipelines.
Exam lens
- Database ≠ warehouse: a general/operational data store versus an analytical, historical, integrated repository.
- Warehouse ≠ mart: enterprise-wide analytical store versus a focused departmental subset.
- Big data ≠ simply “a lot of data”: discuss challenging Vs and the capability needed to derive value.
- Mention the trade-off: warehouses improve consistent analysis but ETL, storage, latency, governance, and cost matter; marts improve focus but can fragment truth.
Rapid revision checklist
- Define OLTP and OLAP in purpose and workload terms.
- Explain warehouse, mart, ETL/ELT, and star schema.
- Recall the 5 Vs with examples.
- Distinguish a data lake from a curated warehouse.
- Explain why common definitions and lineage matter.
Key takeaways
- Operational systems run the business; analytical stores help understand it.
- A warehouse integrates historical data; a mart narrows it for a subject or department.
- Big data is a management and processing challenge across several Vs.
- Analytical value depends on quality, governance, and shared definitions—not volume alone.
Sources
- Rainer & Prince, Management Information Systems (Wiley) — textbook exam framing.
- Laudon & Laudon, Management Information Systems: Managing the Digital Firm, 10th ed. — textbook exam framing for data warehouses and business intelligence.
- Boddy & Boonstra, Managing Information Systems: Strategy and Organization — textbook exam framing.
- IBM, “What is a data warehouse?”, “What is a data mart?”, and “What is big data?” — supplements for definitions and architecture.