1. Introduction
Apache Cassandra is an open-source distributed NoSQL database that is built to handle large amounts of data across multiple data centers. Cassandra’s data model is a topic of discussion across multiple documents and papers, often resulting in confusing or contradictory information. This is due to Cassandra’s ability to store and access column families separately, which results in a mistaken classification as column-oriented rather than column-family.
In this tutorial, we’ll look at the differences between data models and establish the nature of Cassandra’s partitioned row store data model.
2. Database Data Models
The README on the Apache Cassandra git repo states that:
Cassandra is a partitioned row store. Rows are organized into tables with a required primary key.
Partitioning means that Cassandra can distribute your data across multiple machines in an application-transparent matter. Cassandra will automatically repartition as machines are added and removed from the cluster.
Row store means that like relational databases, Cassandra organizes data by rows and columns.
From this, we can conclude that Cassandra is a partitioned rowstore. However, column-family or wide-column are also suitable names, as we’ll find out below.
A column-family data model is not the same as a column-oriented model. A column-family database stores a row with all its column families together, whereas a column-oriented database simply stores data tables by column rather than by row.
2.1. Row-Oriented and Column-Oriented Data Stores
Let’s take an Employees table as an example:
ID Last First Age
1 Cooper James 32
2 Bell Lisa 57
3 Young Joseph 45
A row-oriented database stores the above data as:
1,Cooper,James,32;2,Bell,Lisa,57;3,Young,Joseph,45;
While a column-oriented database stores the data as:
1,2,3;Cooper,Bell,Young;James,Lisa,Joseph;32,57,45;
Cassandra does not store its data like either a row-oriented or a column-oriented database.
2.2. Partitioned Row Store
Cassandra uses a partitioned row store, which means rows contain columns. A column-family database stores data with keys mapped to values and the values grouped into multiple column families.
In a partitioned row store, the Employees data looks like this:
"Employees" : {
row1 : { "ID":1, "Last":"Cooper", "First":"James", "Age":32},
row2 : { "ID":2, "Last":"Bell", "First":"Lisa", "Age":57},
row3 : { "ID":3, "Last":"Young", "First":"Jospeh", "Age":45},
...
}
A partitioned row store has rows that contain columns, yet the number of columns in each row does not have to be the same (like big-table). Some rows may have thousands of columns, while some rows could be limited to just one.
We can think of a partitioned row store as a two-dimensional key-value store, where a row key and a column key are used to access data. To access the smallest unit of data (a column), we must first specify the row name (key) and then the column name.
3. Conclusion
In this article, we have learned that Cassandra’s partitioned row store means that it is column-family rather than column-oriented. The main characteristic that defines column-family is that column information is part of the data. This is the main difference between a column-family model and both row-oriented and column-oriented models. The term wide-column comes from the idea that tables holding an unlimited number of columns are wide by nature.
We’ve also explored how rows in a column-family datastore don’t need to share column names or column numbers. This enables schema-free or semi-structured tables.
