DateTieredCompactionStrategy: Notes From the Field

DateTieredCompactionStrategy (DTCS) is a new compaction strategy in Apache Cassandra designed for time-series data and access patterns. It simplifies the stored data to more closely match the logical view by grouping the stored data for optimal access, which is less intensive than traditional compaction methods. DTCS is best suited for scenarios where data models and access patterns utilize time-ordering for clustering. The strategy can be tuned with parameters such as min_threshold, max_sstable_age_days, and others to balance the compaction workload with the need to optimize read requests to the storage layer. A test showed that DTCS performed well in a time-series workload scenario, maintaining stable latencies and throughput while increasing data density. However, it is crucial to consider operational headroom when using DTCS, as high data density can affect bootstrapping or repairing nodes.