3 ms·

Google shows TridentKV

• derefr 3y ago
  TridentKV is LSM-tree based (i.e. it's "LevelDB in the large" just like BigTable is); it's only "read-optimized" in the sense that it tries to minimize the costs that LevelDB's high merge depth imposes. It still doesn't architecturally make the read-latency/write-commit-latency trade-off in favor of read latency. One could call TridentKV "as read-optimized as a write-optimized KV store can be."