Kakao Applies Distributed Radix Tree to MetaKage to Overcome LSMT Limitations

Kakao's distributed Key-Value storage, MetaKage, implemented a distributed Radix Tree to address the difficulties in supporting Ordered Listing and the Compaction costs inherent in the existing LSMT (Log-Structured Merge-Tree) approach. This structure is built on top of a Hash Table and simplifies the multi-node update problem that occurs during BTree Split operations into two PUT operations, maintaining consistency in distributed environments. Additionally, by utilizing Bloom Filters as memory Hints, it eliminates the I/O bottleneck concentrated on the Root Node during Tree traversal, thereby improving performance in large-scale file storage environments.