Alternative databases and data warehouses

A3.4.1 Types of databases

NoSQL databases

NoSQL databases handle large volumes and many data types, and they store data differently from tables.

{
  "id": "12345",
  "name": "Mali",
  "address": { "street": "Sukhumvit 12", "city": "Bangkok" },
  "services": ["transport", "tourism"]
}

Key:   user:12345
Value: {"name": "Mali", "job": "driver"}

row 12345: name=Mali, email=mali@example.com
row 12346: name=Niran, dateOfBirth=2000-12-12

(Mali) -[FOLLOWS]-> (Niran)
(Niran) -[LIKES]-> (Post 9)
(Mali) -[COMMENTED_ON]-> (Post 9)

Other database types

A self-managed cloud database gives you full control of the database, while the provider runs the infrastructure. A managed service has the vendor handle backups, scaling and patching.

Next: data warehouses →

A3.4.2 Data warehouses

A data warehouse stores large amounts of historical data from many sources for analysis, not for day-to-day transactions.

Taking care with ETL

• Extraction must not disrupt the source system.
• Different formats must be converted into something readable.
• Extracted data must be relevant to what users need.

Next: OLAP and data mining →

A3.4.3 OLAP and data mining

Business intelligence uses technology to collect, analyse and present information for better decisions.

Online analytical processing (OLAP)

OLAP lets users interactively explore data across dimensions such as time, product and region. Summaries are pre-aggregated, so they come back quickly and can feed charts and dashboards.

Data mining

Data mining finds hidden patterns in large data sets using statistics and machine learning, for forecasting, customer segmentation and fraud detection.

Next: distributed databases →

A3.4.4 Distributed databases

A distributed database keeps its data on different sites, yet users see one unified database. Online shops and telecoms companies use them across regions.

ACID across nodes

Features

Choose a run.

Next: review →

Quick quiz

Exam-style questions

These are written for this site in the style of Paper 1. Answer on paper first, then open the mark points.

1. Define the term “data mining”. [1]

   Mark points

   • The process of sorting through large data sets to find patterns and relationships that help solve problems [1].

2. Explain the role of integrated data in a data warehouse. [2]

   Mark points

   • Data from many systems is combined using ETL [1].
   • This gives one consistent source of truth for analysis [1].

3. Describe concurrency control in distributed databases and how it can be achieved. [4]

   Mark points

   • It manages simultaneous access to shared data on several nodes [1].
   • It prevents conflicts such as lost updates [1].
   • Locks stop other transactions using data until commit [1].
   • Or timestamps set the order in which transactions run [1].

4. Compare a document database with a relational database. [4]

   Mark points

   • A relational database uses a fixed schema of tables, while documents are flexible [1].
   • Documents suit unstructured or hierarchical data [1].
   • Relational data has less redundancy, because of normalization [1].
   • Relational databases give strong transactions and joins, while document stores scale out more easily [1].

5. Compare classification and sequential pattern discovery. [4]

   Mark points

   • Classification assigns data to known categories [1].
   • It is used for spam detection, for example [1].
   • Sequential pattern discovery finds events in a time order [1].
   • It is used for clickstream analysis, for example [1].

6. Outline the role of OLAP in business intelligence. [3]

   Mark points

   • It lets users analyse data across dimensions interactively [1].
   • Pre-aggregated views allow quick summaries [1].
   • Ad hoc queries answer questions without predefined reports, which supports decisions [1].

7. Outline two differences between a database view and a data warehouse. [2]

   Mark points

   • A view is a virtual table from a query, while a warehouse physically stores data [1].
   • A view presents existing data from one database, while a warehouse combines historical data from many sources [1].