Clustrix uses layered trees for table/index storage. Without going into a great amount of detail, this means that all writes are done to a top-layer tree, and later merged down into lower layers in the background. These merge operations occur periodically based on the write volume of the table/index in question, and are transparent to users. Further, the merges produce lower layers which are compact b-trees with no fragmentation.
Have more questions? Submit a request