Database and DBMS Basics, and Types of NoSQL

A database (データベース) is a collection of data organized for a purpose and stored so that many users and applications can share it.

If each department keeps the same customer details in its own separate files, one copy may get a new address while the other does not, and the two disagree. Gathering the data in one place and organizing it cuts duplication and lets everyone see the same up-to-date content.

AnalogyIt is like keeping one shared address book in the living room instead of each family member writing contacts in a personal notebook. When someone updates it, everyone sees the same entries.

How to rememberRead the word literally: the "base" where data is gathered and kept.

1 past questions on this term データベース →

A DBMS (Database Management System) is software that manages a database and reads and writes data on behalf of users and applications.

The application does not deal with where or in what format data is stored; it asks the DBMS to fetch or change data. The DBMS handles concurrency control (同時実行制御), coordinating operations, for example by making some wait, so that data stays consistent even when many people update it at once. It also handles failure recovery (障害回復), which restores data damaged by a failure from backups and the like, and access rights management (アクセス権管理), which sets what each user may do.

AnalogyIt is like a library's stacks clerk. Patrons do not enter the stacks but ask at the counter; the clerk finds the book, hands it over, and keeps the loan and return records.

How to rememberManagement means managing: the DBMS carries "data manager" right in its name.

3 past questions on this term DBMS →

An RDBMS (Relational Database Management System) is a DBMS that manages a database based on the relational data model (関係データモデル), which holds data in the form of tables.

Each row of a table is one item of data (a record) and each column is one field, with column names and data types fixed in advance. Giving two tables a field in common, such as the product number, lets you link them and retrieve data across both. An RDBMS is operated with a language called SQL.

AnalogyIt is like setting up several spreadsheet sheets with the column headings decided first, then matching rows across them by a shared number.

How to rememberR stands for Relational. In the relational data model a table itself is called a "relation," so read it as "a DBMS that manages data in tables."

2 past questions on this term RDBMS →

NoSQL is an umbrella term for databases that hold data in forms other than the tables of the relational data model.

The name is usually read as "Not only SQL," and it refers not to one product but to several types that hold data in different ways. Many of these types do not require table columns to be fixed in advance, and they are used for data whose shape varies or for large volumes of data spread across many servers.

AnalogyIf an RDBMS is manuscript paper with a fixed grid, NoSQL is a stationery shelf holding plain notebooks, flashcards, sticky notes and more.

How to rememberNo is usually read not as "no" but as "Not only." Read that way, the name does not reject SQL; it admits ways of holding data other than tables.

A key-value store (キーバリューストア) is a NoSQL type that saves a key, used to look up data, together with its content, the value, as one pair.

Specify a key and the one matching value comes back; the mechanism is that simple, so reads and writes are fast. Because the DBMS does not define the form of what is inside a value, it is a poor fit for searches that use a field inside the value as the condition.

AnalogyIt is like a cloakroom that works with numbered tags. Give the tag number and your item comes right out, but you cannot ask the staff to find it by what is inside.

How to rememberIt has the same shape as a dictionary entry: headword, then definition. The headword is the key and the definition is the value.

KVS is short for key-value store, formed from the initials of Key-Value Store.

It means the same thing as a key-value store, and texts use either the Japanese name (キーバリューストア) or the abbreviation. Its trait of holding data as key-value pairs applies just the same.

How to rememberExpand K (Key), V (Value) and S (Store) in order and you have the full name.

A document-oriented database (ドキュメント指向データベース) is a NoSQL type that saves each document that has fields, such as JSON or XML, as one record.

Each document may carry different fields, so there is no need to align columns as in a table. Because the DBMS can read the fields inside a document, you can also search using a field's value as the condition.

AnalogyRather than a fixed application form, it is like a binder of free-form cards, each with different boxes filled in depending on the person.

How to rememberA document is a piece of paperwork: one sheet is one record, just as it is.

A graph-oriented database (グラフ指向データベース) is a NoSQL type that saves data as nodes (points) and the connections between data as edges (lines).

← Scroll the figure sideways →

"Graph" here does not mean a line chart but a network of points joined by lines. Because it stores the connections themselves as data, it is good at following links through many steps, such as friends of friends or routes from station to station.

AnalogyIt is like saving a relationship chart as is, with each person's name circled and lines drawn between people who are connected.

How to rememberA graph here means not bars or lines on a chart but a rail map of stations (points) and tracks (lines).