Database Reference
In-Depth Information
Write performance
Cassandra writes are sequential; all it needs to do is append it to the commit log and put it
in memory. There is not much that can be done internally to Cassandra's setting to tweak
writes. However, if disk writes are fast, and somehow I/O contentions can be lowered due
to multiple things that happen in the Cassandra life cycle, such as flushing MemTable to
disk, compactions, and writes to commit logs, it can boost the write performance.
So, having fast disks, commit logs, and data files in separate dedicated hard disks directly
attached to the system, will improve write throughput.
Search WWH ::




Custom Search