When the use case aligns with the architectural limitations, Cassandra excels at storing and accessing datasets up to petabytes in volume, delivering impressive throughput. As the data or workload volume grows, we expand the cluster linearly, ensuring consistent performance.
However, even when we adhere to the documentation and best practices and create an effective data model, we might encounter underperforming nodes or unexpected challenges with throughput scaling after a cluster expansion—and it’s not always clear what causes the imbalance. Linear scalability relies on the assumption that workload and data are evenly distributed across all nodes in a cluster, and the cluster capacity relates directly to the number of nodes. Sometimes, these conditions aren’t met, affecting linear scalability. So, we strive for scalability and balance and are willing to fulfill the necessary conditions.
Read on for a few common performance issues and what you can do about them.