Volume, Velocity and Variety

Overview

Recall questions on what Big Data means, the three characteristics and the test for each, structured against unstructured data, and the sources that stream continuously.

What you need to know

Answer each one before you reveal it.

Questions from this subtopic

Each one with the model answer underneath.

1. 1

   What does the term Big Data actually refer to?

   Model answer

   It is a loose label rather than a precise measurement. It covers any collection of data that has outgrown the ordinary tools an organisation would normally reach for, so that the usual way of storing it and asking questions of it stops working. There is no fixed number of gigabytes at which data becomes Big Data. The point is that the containers it used to sit in no longer hold it, and the methods that used to interrogate it no longer finish in a useful amount of time.

2. 2

   What are the three characteristics conventionally used to describe Big Data?

   Model answer

   Volume, velocity and variety. Volume is the sheer quantity of it. Velocity is the rate at which it arrives and the speed at which an answer is wanted. Variety is the mixture of different forms the data takes. A dataset does not have to be extreme in all three to qualify — being unmanageable on any one of them is usually enough to force a change of approach.

3. 3

   What does volume mean, and what is the practical test for it?

   Model answer

   Volume is the quantity of data. The practical test is whether it will still fit on one machine: once a dataset is larger than a single server can hold or work through, volume has become the limiting factor. That threshold matters more than any absolute figure, because it is the point at which the data must be spread across several machines and every tool that assumed one machine has to be replaced.

4. 4

   What does velocity mean, and what response times does it imply?

   Model answer

   Velocity is the rate at which data arrives and has to be dealt with. Rather than sitting still in a store waiting to be queried overnight, the data streams in continuously and a response is expected in milliseconds to seconds. Fraud checks on card transactions and alerts from sensor readings are typical: an answer produced an hour later has no value, because the situation it described has already passed.

5. 5

   What does variety mean, and what kinds of data does it cover?

   Model answer

   Variety is the range of different forms the data takes. Alongside neat tabular records there is free text, email, images, audio, video, and log files, none of which share a common layout. Some of it is structured, some has partial structure such as tags or headers, and much of it has none at all. The collection as a whole therefore has no single shape that one storage format or one query language can assume.

6. 6

   Why is 'big' described as a relative term rather than a fixed threshold?

   Model answer

   Because what overwhelms one organisation is routine for another. A volume that defeats a small firm's single server is unremarkable to a company running a large cluster, and a quantity considered enormous a decade ago is now handled on a desktop. What makes size matter is not the figure itself but the moment the data stops fitting on one machine, since that is when the familiar tools stop being an option.

Other ways to work this

The same content, four other ways.

What is inside Levo

Six ways to work the same material.