MODULE 01 · FOUNDATIONS

Big data isn't about size — it's about what your tools can no longer ignore.

This is a click-through primer, not a slideshow. Every card, stage, and layer below responds to a click — explore them in any order to build the full picture.

Start with the three Vs below

The three Vs

Before "big data" meant a tech stack, it meant a problem definition. Click each letter to see why traditional databases struggle with it.

V₁
Volume
tap to expand →

The sheer scale of data generated — often too large to fit on a single machine's disk or process in memory at once.

e.g. 500 hours of video uploaded to YouTube every minute
V₂
Velocity
tap to expand →

The speed at which new data arrives and must be acted on — sometimes in real time, not in nightly batches.

e.g. stock ticks, sensor streams, fraud detection at swipe-time
V₃
Variety
tap to expand →

The mix of structured, semi-structured, and unstructured formats — tables, JSON logs, images, free text, audio.

e.g. a single pipeline ingesting CSVs, tweets, and CCTV footage

The analytics pipeline

Raw data becomes a decision through four stages. Click a stage to see what actually happens there and which tools are typically used.

STAGE 01

Collect

Pull data from its source

STAGE 02

Store

Land it somewhere durable

STAGE 03

Process

Clean, join, transform

STAGE 04

Analyze

Query, model, visualize

STAGE 01 · COLLECT

Collecting raw data

Data enters the system from APIs, application logs, IoT sensors, or third-party feeds — often continuously, not as a one-time export.

A typical big data stack

Most real systems are layered — each layer trades off durability, speed, and cost differently. Click a layer to expand it.

Ingestion Layer

+

Brings data in from outside the system, often as continuous streams. Tools like Kafka or Kinesis buffer incoming events so downstream systems aren't overwhelmed by spikes.

Storage Layer

+

Holds raw and processed data cheaply at scale — distributed file systems (HDFS), object storage (S3), or data lakes that don't require a fixed schema up front.

Processing Layer

+

Transforms raw data into something usable — distributed engines like Spark or Flink run computations across many machines in parallel.

Serving Layer

+

Exposes results to people and applications — dashboards, BI tools, or APIs that let analysts and downstream systems query the processed data.

Quick check

One question to confirm the idea stuck.

A hospital wants to flag abnormal heart-rate readings from wearable devices the instant they happen, rather than reviewing them at the end of the day. Which "V" is this primarily about?
Volume — there's a lot of patient data
Velocity — the data must be acted on immediately
Variety — wearables produce many data formats