Performance

Performance comparison between snmpkit (Rust core) and pysnmp (pure Python).

Benchmarks

Analysis

BER Encode: 127x faster

The most critical operation for request/response performance. snmpkit encodes SNMP messages in Rust with zero-copy byte handling, while pysnmp uses Python’s pyasn1 library with multiple object allocations.

GET Request: 8.5x faster

End-to-end GET request latency including:

• Message encoding
• UDP send/receive
• Response decoding

The speedup is lower than BER encoding alone because network I/O dominates. The 8.5x improvement comes from faster encoding/decoding on both ends.

WALK: 14x faster

Walking an OID subtree shows the cumulative benefit:

• Multiple GETBULK requests
• Response decoding for each batch
• Iterator overhead

snmpkit’s bulk_walk is particularly efficient because the Rust decoder processes batches without Python object creation until values are accessed.

Test Environment

Real-World Impact

Polling Throughput

For network monitoring polling 1000 devices:

Concurrent Requests

With asyncio, snmpkit can handle more concurrent requests before CPU becomes the bottleneck:

async def poll_many(hosts: list[str]):
    async with asyncio.TaskGroup() as tg:
        for host in hosts:
            tg.create_task(poll_device(host))

Memory Usage

snmpkit uses less memory per OID:

• pysnmp: Full Python objects for each OID component and value
• snmpkit: Compact Rust types, Python wrappers created on-demand

Reproducing Benchmarks

Run the benchmarks yourself:

# Install snmpd for local testing
sudo apt install snmpd
sudo systemctl start snmpd
 
# Run benchmarks
uv run python -m benchmarks.run_all

Results are saved to benchmarks/results.json.

Individual Benchmarks

# BER encoding only (no network)
uv run python benchmarks/bench_ber.py
 
# GET requests
uv run python benchmarks/bench_get.py
 
# WALK operations
uv run python benchmarks/bench_walk.py

Custom Target

SNMP_HOST=192.168.1.1 SNMP_COMMUNITY=private uv run python benchmarks/bench_get.py