One aspect to consider when planning Cassandra node instances is to take the approach of many, smaller instances versus fewer, larger instances. Each approach has its share of advantages and shortcomings.
Larger nodes, consisting of 24+ core CPUs, 64 GB of RAM or more, and disk space consisting of more than a TB, can offer the benefit of concentrating resources onto a smaller number of nodes. Depending on the resource parameters chosen, they can also cost less in the cloud.
But larger nodes can sometimes be difficult to manage in terms of streaming operations, such as repairs or joining new nodes or data centers. Larger nodes also typically means larger JVM heaps, which (prior to ...