Topology

Before the rise of big tech like Facebook, Apple, and Instagram, relational database management systems (RDBMS) dominated the database market for decades. The advent of big data made the global datasphere grow 15-fold. RDBMS could not cater to the new, humongous data volume, nor could they meet the performance requirements it posed. This gave rise to the development of NoSQL databases like Cassandra. Cassandra has addressed and successfully resolved the constraints of previous data management technologies.

Tech giants have adopted Cassandra as it provides continuous availability, simplified operations, linear scalability, and effortless automatic distribution of data across multiple datacenters and cloud availability zones.

In 2014, Apple alone ran 75,000 Cassandra instances, storing more than 10 petabytes of data in a single cluster with more than a thousand nodes. Today, Apple has more than 160,000 Cassandra instances storing 100 petabytes of data across more than a thousand clusters.

Similarly, Netflix runs more than 10,000 Cassandra instances, storing 6 petabytes of data across more than a hundred clusters catering to over 1 trillion requests per day. 

Let’s look at the fundamentals of Apache Cassandra. Each topic will cover the theory followed by hands-on experience using Cassandra query language (CQL).

Cassandra’s cluster topology

Cassandra can be conceptualized as a collection of nodes organized in a ring. Nodes in a ring constitute a datacenter, and a Cassandra cluster can span multiple datacenters. 

Cassandra is a peer-to-peer, leaderless system in which all nodes have the same role. The absence of master/slave/replica sets eliminates a single point of failure, simplifies management, and provides load balancing, reliability, and scalability.

Data is distributed and replicated across multiple nodes (Cassandra installations) within clusters.