📄 Paper 2 · 4.11 Big Data
✓ Free lesson
4.11.1 Big Data
AQA 7517 · A-Level Computer Science · ~15 min read

What is Big Data?

Big data refers to datasets that are so large, fast-growing, or complex that traditional database management systems cannot process them efficiently. Big data is characterised by the 3 Vs:

Volume — Vast quantitiesThe scale of data is enormous — terabytes to petabytes or more. Examples: all Twitter posts ever made, NHS patient records, Google search logs, CCTV footage.
Velocity — High speed of generationData arrives at very high speed and must be processed rapidly (often in real time). Examples: stock market ticks, sensor data from IoT devices, social media streams, fraud detection.
Variety — Many different formatsData comes in many formats: structured (SQL tables), semi-structured (JSON, XML), and unstructured (video, images, social media text, audio). Traditional RDBMS handle only structured data well.

Traditional vs Big Data Processing

AspectTraditional DBBig Data
Data sizeGB rangeTB–PB range
Data typesStructured onlyStructured + semi + unstructured
Processing modelCentralisedDistributed (across many machines)
SchemaFixed schema before dataSchema on read (flexible)
Update modelFrequent CRUD updatesBatch / stream processing

Distributed Processing — Hadoop and MapReduce

Because big data cannot fit on or be processed by a single machine, it requires distributed processing — spreading work across many computers (a cluster).

Apache Hadoop

Hadoop is an open-source distributed processing framework. It uses two core components:

  • HDFS (Hadoop Distributed File System): stores data split into blocks across many nodes. Each block is replicated (typically 3 times) for fault tolerance.
  • MapReduce: programming model for processing data in parallel across the cluster.

MapReduce Model

  • Map phase: each node processes its local data in parallel, producing key-value pairs. Example: count occurrences of each word in documents → (word, 1) pairs.
  • Reduce phase: key-value pairs with the same key are aggregated. Example: all ("hello", 1) pairs → ("hello", 42).

MapReduce enables processing of massive datasets by splitting the work into small parallel tasks, then combining results.

Machine Learning with Big Data

Large datasets enable machine learning algorithms to find patterns and make predictions that would be impossible with smaller datasets:

  • Recommendation systems (Netflix, Amazon) — analyse millions of user interactions
  • Fraud detection — identify unusual transaction patterns in real time
  • Medical diagnosis — train models on millions of patient records
  • Natural language processing — large language models trained on internet-scale text

Ethical Issues with Big Data

  • Privacy: data collected without explicit consent; individuals may not know what data is held about them
  • Bias: biased training data leads to biased models (e.g. facial recognition failing on darker skin tones)
  • Security: large repositories are high-value targets for breaches
  • Data ownership: who owns data collected about individuals?
  • Surveillance: mass collection can enable government or corporate surveillance
Exam tip: AQA 7517 requires the 3 Vs (Volume, Velocity, Variety) with examples. Know why traditional RDBMS struggle (centralised, structured-only, fixed schema). Understand MapReduce conceptually — Map (parallel processing, key-value pairs), Reduce (aggregation). Also know at least 2–3 ethical issues. Big data questions often appear as 4–6 mark essay-style answers.
Click through the slides at your own pace. Use arrow keys or click to advance.
Click slide or press arrow keys to navigate

Worksheet — 4.11.1 Big Data

8 questions · instantly marked · AQA 7517 standard

Q1State and explain the three Vs that characterise big data.[4]
✅ Mark scheme
Mark scheme
Volume: enormous scale of data — terabytes to petabytes; too large for traditional systems [1]; example: all social media posts, satellite imagery [1]. Velocity: data is generated at very high speed and must be processed rapidly — often in real time [1]; example: stock market trades, sensor streams, social media feeds [1]. Variety: data arrives in many formats — structured (tables), semi-structured (JSON/XML), and unstructured (video, images, audio, text) [1]; traditional RDBMS struggle with unstructured formats [1].
Q2Explain why traditional database management systems (RDBMS) are not suitable for handling big data.[2]
✅ Mark scheme
Mark scheme
Centralised architecture — cannot scale to petabyte-scale storage/processing on a single machine [1]; only handle structured data well — cannot process unstructured types like video or social media text [1]; fixed schema defined before data is loaded — big data often has variable or unknown structure [1]; batch-oriented update model is too slow for real-time high-velocity streams [1].
Q3Describe what is meant by distributed processing and explain why it is needed for big data.[2]
✅ Mark scheme
Mark scheme
Distributed processing: spreading data and computation across many computers (a cluster) working in parallel [1]; each node processes a portion of the data simultaneously [1]. Needed because: no single machine has sufficient storage or processing power for petabyte-scale datasets [1]; distributing work allows linear scaling — adding more nodes increases capacity [1].
Q4Explain the MapReduce model. Include what happens in both the Map and Reduce phases.[2]
✅ Mark scheme
Mark scheme
Map phase: data is split across nodes; each node processes its local portion in parallel and outputs key-value pairs [1]; example: word count — each document produces (word, 1) pairs [1]. Reduce phase: all key-value pairs with the same key are collected and aggregated [1]; example: all ("hello", 1) pairs are summed → ("hello", 42) — the total count [1].
Q5Describe two roles of HDFS (Hadoop Distributed File System) in the Hadoop framework.[2]
✅ Mark scheme
Mark scheme
Data storage: HDFS stores data split into fixed-size blocks distributed across many nodes in the cluster [1]; enables parallel access — nodes can read their local data during the Map phase [1]. Fault tolerance: each block is replicated across multiple nodes (default: 3 copies) so data is not lost if a node fails [1]; if one node fails, another copy is available — HDFS automatically re-replicates the lost block [1].
Q6Describe two ways machine learning benefits from big data.[2]
✅ Mark scheme
Mark scheme
Any 2 of (1 mark for description + 1 mark for example each): better accuracy — more training data reduces overfitting and improves generalisation; e.g. image classifiers trained on millions of images are far more accurate [1+1]; pattern detection — large datasets reveal patterns invisible in smaller samples; e.g. fraud detection identifies unusual transaction clusters [1+1]; personalisation — recommendation systems trained on millions of user interactions accurately predict preferences; e.g. Netflix, Spotify [1+1]; NLP — language models require internet-scale text data to learn grammar, context, and meaning [1+1].
Q7Discuss two ethical issues raised by the collection and use of big data.[2]
✅ Mark scheme
Mark scheme
Any 2 of (1+1 per issue — description + impact): Privacy: personal data collected without explicit consent — individuals unaware of what is held and how it is used [1]; impact: erosion of personal autonomy and trust [1]. Bias: if training data reflects historical biases, models perpetuate them; e.g. facial recognition accuracy varies by skin tone [1]; impact: discriminatory outcomes in healthcare, hiring, policing [1]. Security: large centralised data stores are high-value breach targets [1]; impact: mass exposure of personal information (e.g. healthcare records) [1]. Surveillance: governments/corporations can track individuals at scale [1]; impact: chilling effect on free expression and behaviour [1].
Q8A company collects 500 TB of social media posts, CCTV video, and transaction records daily. Explain why big data techniques are needed and suggest two specific technologies or approaches they could use.[4]
✅ Mark scheme
Mark scheme
Why needed: 500 TB per day exceeds the capacity of any single RDBMS — too large for centralised storage/processing [1]; variety of data types (text posts, video, structured transactions) cannot all be handled by a relational schema [1]; high velocity — daily 500 TB requires rapid ingestion and real-time or near-real-time processing [1]. Technologies (2 marks each — name + purpose): Hadoop/HDFS — distributed storage splits data across many nodes; MapReduce processes each type in parallel [1+1]; Spark — in-memory distributed processing framework, faster than MapReduce for iterative analysis [1+1]; NoSQL databases (e.g. MongoDB, Cassandra) — store unstructured/semi-structured data without fixed schema [1+1]; stream processing (e.g. Apache Kafka) — handles real-time high-velocity data ingestion [1+1].
Big Data Quiz
Question 1 of 15
You scored
out of 15
Card 1 of 8
Click to reveal definition
🎉
All cards reviewed!
TermDefinition
🎯

Mini Test — Big Data

10 questions · 10 minutes

← 4.10.5 Transactions
68 of 70 · AQA 7517
4.12.1a Functional Programming →