Filter Optimization

Application scenario: There are certain bottlenecks in searching for data in databases.

Technical principle: During the flush of in-memory immutable MemTables to drives and the subsequent compaction and deduplication of SST files, more bitmap bits are allocated to hot data to improve the accuracy of hot data identification and reduce the overhead of data search.

Performance metric: The performance in YCSB workload A and workload C scenarios is improved by 5% on average.

How to use:

For details, see RocksDB 6.1.2 Filter Optimization Feature Guide.