SecantusDB Rust server

The SecantusDB Rust server is a whole second implementation of the SecantusDB database — wire parsing, command dispatch, cursors, change streams, the operator engines, and WiredTiger storage, all in Rust with no Python in the request path. It speaks the same MongoDB wire protocol as the Python server, passes the same unmodified driver-conformance suites (99.5% of pymongo’s own tests — level with the Python server), and ships as a single static-WiredTiger binary: secantusd-rs.

# From a prebuilt archive (Linux x86_64 / macOS arm64):
./secantusd-rs --port 27017 --storage-path ./secantus-data

# Or bundled in the Python wheel:
#   pip install SecantusDB   (storage-engine build)
import _secantus_server
from pymongo import MongoClient

srv = _secantus_server.RustServer("./secantus-data", 0)  # port 0 = OS-assigned
host, port = srv.address
client = MongoClient(host, port, directConnection=True)
client["mydb"]["users"].insert_one({"_id": 1, "name": "Joe"})
srv.stop()

Every MongoDB driver that talks to the Python server talks to the Rust server unchanged — same OP_MSG handshake, same commands, same error codes, same on-disk WiredTiger semantics.

And it is fast: on the nine-workload benchmark the Rust server runs at ~0.7×–2.2× of real mongod per operation (delete, $group, and the change-stream drain all beat mongod at 0.7–0.9×, update near parity — after a mimalloc allocator, LTO, and profile-guided optimization cut the BSON-materialization allocation and hot-path branch cost), sustains ~2.5× multi-writer scaling fully durable (monotonic to eight writers), and is roughly 1.9×–18× faster than the Python server workload-for-workload — measured end-to-end through pymongo on on-disk WiredTiger. Numbers and methodology: Benchmark and Concurrency.

Where this fits

SecantusDB ships two separate servers on independent version lines. The Python server is the conformance reference and the default choice; the Rust server is the same database built for speed and dependency-free standalone deployment. The full decision guide is The two servers, and the feature comparison maps both servers against real MongoDB, feature by feature.

Contents