Two systems, one interface
The local database knows nothing about the network; the Raft layer replicates opaque commands and knows nothing about the LSM tree.
DATABASE INTERNALS · DISTRIBUTED SYSTEMS
TinyKV Distributed Store
I built a durable C++ key-value store that combines an LSM engine, key-value separation, crash recovery, and Raft consensus behind a small API.
33,800 replicated writes/s · 8 clients
12× faster writes for 16 KiB values
65× throughput gain from group commit
60 deterministic GoogleTest cases
Crash safety by construction
The engine avoids in-place data updates and treats the manifest as the durable description of every live file.
Raft and measured trade-offs
A three-node cluster acknowledges writes only after a majority has them on disk, while deterministic tests inject partitions and divergent logs without timing races.