Given its log-based storage model and memtable/commitlog design, Apache Cassandra usually does very well when it comes to high-performance data-ingestion. That being said, it is not uncommon to hear of application teams complaining about Cassandra nodes being saturated with load due to writes. Here is a short list of things to check when write performance suffers.
Write performance
Commitlog mount point
One piece of advice, going back to the early days of Apache Cassandra, was to ensure that the commitlog was on a different physical mount point than the data drives. This is because disk I/O could become bottlenecked at the device level during periods of heavy writes. Putting the commitlog and data directories on separate mount...
