Data Distribution in Cassandra

Data distribution is done through consistent hashing to make the data evenly distributed across the nodes in a cluster. Instead of getting rows of table on a single node, rows gets distributed across the clusters which make the load of table data get evenly divided. The partition key is used to distribute data among nodes and determine data locality.

In Cassandra, data distribution and replication work together. Mainly depends on three things i. e. partition key, key value, and token range.

Cassandra Table

This table consists of two rows in which one row has four columns followed by their values whereas the second row contains two columns with their values present. column 1 is having the primary key.

Example

In this example, we are going to show how the data is distributed over a cluster.

id     name      dob                     salary
1       x        5−12−2000                60000            
2       y        6−12−1999                70000         
3       z        7−12−1998                80000               
4       a        8−1−1998                 90000     

Below given ring architecture has four nodes provided with the token range and a specific token id. The partitioner will generate token values and assign them over clusters.

Partitioner

It helps in data distribution across the nodes in a Cassandra cluster. Hashing is used on the partition key of row data to get the token values. The default partitioner in Cassandra is Murmur3.

Token

They are hash values that are used to store rows in the ring. The token is a 64−bit integer. The range for the token is −2^63 to 2^63.

For example: Let's take some random hash values for the above table.

PARTITION KEY	MURMUR 3 HASH VALUE
 x	              −700
 y	              −300
 z	               400
 a	               700

Replication factor

It shows the total number of copies of data across the cluster. Replication factor=1 means we have only one copy of data so if it goes down, there is no way to recover it.

The replication factor=2 means two copies are present and so on.

Conclusion

This article consists of data distribution in Cassandra. Data is distributed through hashing in the form of clusters. The main components are the partition key, key value, and token. The partition key performs the distribution across the nodes in clusters. Tokens are used to store rows in ring form. The replication factor decides the number of copies of the data across the cluster is made.