WhizMQ - A message queue library implementing ZMTP 3.0 in Nim.
Built on top of PowPow event library

nimble install whiz

API reference

😍 Key Features

• ZMTP 3.0 wire protocol, full greeting, framing, and command handling
• Socket patterns: PAIR, PUB/SUB, REQ/REP, PUSH/PULL, CLIENT/SERVER, File Transfer
• Security mechanisms:
  • NULL: no authentication (plaintext)
  • PLAIN: username/password authentication with ZAP callback
  • CURVE: X25519 + XChaCha20-Poly1305 AEAD encryption and mutual authentication
• Transport TCP (with optional TLS on POSIX), IPC (Unix domain sockets), and WebSocket (RFC 6455)
• Built on PowPow event notification library in Nim

Note

CURVE is powered by nimcypher, a 100% pure-Nim port of Monocypher 4.0.3 — no C compiler flags, no system libraries to install.

🗺 Roadmap

• ZMTP 3.0 wire protocol (greeting, framing, commands)
• PAIR, PUB/SUB, REQ/REP, PUSH/PULL socket patterns
• File transfer over ZMTP
• PLAIN security mechanism + ZAP auth callbacks
• CURVE security mechanism (X25519 + AEAD)
• RADIO/DISH (dgram) socket pattern
• CLIENT/SERVER (stream) socket pattern
• CURVE vouch (Ed25519 signature in HELLO for key continuity)
• TLS/DTLS transport
• GSSAPI / Kerberos mechanism
• WebSocket transport (RFC 6455)
• Formal ZMTP conformance tests
• Performance benchmarks vs ØMQ

Examples

...

Benchmarks

Note

Benchmark results are not consistent and may show different results across runs and environments.

Column legend:

• benchmark — test name: {pattern}_{scenario}_{transport} (_tcp or _unix)
• subscriber — specific subscriber/worker for multi-endpoint tests; total is the aggregate row, - for single-endpoint
• n — number of messages sent (throughput) or round-trips (latency)
• size — payload size in bytes
• total(μs) — wall-clock duration of the test in microseconds
• throughput/s — messages per second = n / total(μs) * 1_000_000
• min(μs) — fastest observed per-message time (minimum)
• p50(μs) — median per-message time (50th percentile)
• p75(μs) — 75th percentile per-message time
• p90(μs) — 90th percentile per-message time
• p99(μs) — 99th percentile per-message time
• avg(μs) — arithmetic mean per-message time
• σ(μs) — population standard deviation of per-message times

Per-message time collection:

• Latency benchmarks (pair_latency, reqrep): each value is one request–reply round-trip measured from the client
• Throughput benchmarks (all others): inter-arrival deltas between consecutive receives — approximates per-message service time distribution
• bench_sizes_pair uses ACK-windowed flow control: the PAIR server echoes a 1-byte ack every 250 messages and the client keeps at most 250 unacked messages in flight, so bulk transfers measure delivered throughput instead of buffer absorption
• Every benchmark arms a watchdog timer: if the run stalls, results are discarded with a diagnostic instead of hanging the suite

benchmark	subscriber	n	size	total(μs)	throughput/s	min(μs)	p50(μs)	p75(μs)	p90(μs)	p99(μs)	avg(μs)	σ(μs)
bench_pair_latency_tcp	-	5000	1	413222	12100	73.7	76.1	80.9	100.0	145.3	82.5	15.0
bench_pair_latency_unix	-	5000	1	372268	13431	67.8	68.4	73.6	89.5	125.7	74.3	12.7
bench_pair_throughput_tcp	-	10000	512	27907	358332	0.2	0.2	0.2	2.0	5.0	0.7	6.1
bench_pair_throughput_unix	-	10000	512	64441	155179	0.2	0.2	0.2	1.9	82.5	4.9	17.6
bench_pubsub_1sub_tcp	-	10000	512	23566	424336	0.2	0.2	0.3	2.0	4.8	0.7	6.2
bench_pubsub_1sub_unix	-	10000	512	22232	449790	0.2	0.2	0.2	1.8	10.1	0.9	2.5
bench_pubsub_2sub_tcp	sub0	5000	256	23489	212864	0.2	0.2	0.2	0.2	7.8	0.8	24.1
bench_pubsub_2sub_tcp	sub1	4143	256	23489	176379	0.2	0.2	0.2	0.2	7.8	0.8	20.5
bench_pubsub_2sub_tcp	total	9143	256	23489	389244	0.2	0.2	0.2	0.2	7.8	0.8	22.5
bench_pubsub_2sub_unix	sub0	5000	256	14531	344082	0.2	0.2	0.3	0.3	35.0	1.3	5.7
bench_pubsub_2sub_unix	sub1	5000	256	14531	344082	0.2	0.2	0.3	0.3	35.0	1.3	5.7
bench_pubsub_2sub_unix	total	10000	256	14531	688164	0.2	0.2	0.3	0.3	35.0	1.3	5.7
bench_reqrep_tcp	-	5000	128	450034	11110	73.7	78.5	96.3	120.5	164.8	89.8	21.4
bench_reqrep_unix	-	5000	128	367558	13603	68.0	68.6	71.5	86.8	116.7	73.4	10.5
bench_sizes_pair_tcp	-	2000	64	7525	265775	0.1	0.2	0.2	0.2	4.0	3.8	56.6
bench_sizes_pair_tcp	-	2000	1024	9360	213652	0.2	0.2	2.3	2.7	5.0	4.7	60.0
bench_sizes_pair_tcp	-	2000	65536	282384	7082	24.5	28.1	33.2	44.8	351.2	141.2	1584.9
bench_sizes_pair_unix	-	2000	64	4612	433560	0.1	0.2	0.2	0.3	70.6	2.3	23.0
bench_sizes_pair_unix	-	2000	1024	23033	86830	0.2	0.2	1.8	66.0	91.0	11.5	42.2
bench_sizes_pair_unix	-	2000	65536	1372650	1457	538.6	550.1	594.5	665.8	826.3	686.3	1691.0
bench_pushpull_throughput_tcp	-	10000	512	19621	509656	0.2	0.2	0.2	1.9	5.0	0.6	5.7
bench_pushpull_throughput_unix	-	10000	512	60745	164620	0.2	0.2	0.2	1.9	82.0	4.9	17.8
bench_pushpull_1worker_tcp	-	10000	128	9022	1108394	0.1	0.2	0.2	0.2	2.2	0.3	3.2
bench_pushpull_1worker_unix	-	10000	128	7878	1269208	0.1	0.2	0.2	0.2	8.8	0.4	1.3
bench_pushpull_3workers_tcp	worker0	1667	128	10084	165302	0.2	0.2	0.2	0.2	8.2	0.8	17.4
bench_pushpull_3workers_tcp	worker1	1667	128	10084	165302	0.2	0.2	0.2	0.2	4.2	0.8	20.7
bench_pushpull_3workers_tcp	worker2	1666	128	10084	165203	0.2	0.2	0.2	0.2	4.1	0.9	22.0
bench_pushpull_3workers_tcp	total	5000	128	10084	495807	0.2	0.2	0.2	0.2	7.6	0.8	20.1
bench_pushpull_3workers_unix	worker0	1667	128	4079	408632	0.2	0.2	0.2	0.2	46.7	1.0	6.5
bench_pushpull_3workers_unix	worker1	1667	128	4079	408632	0.2	0.2	0.2	0.2	46.6	1.0	6.3
bench_pushpull_3workers_unix	worker2	1666	128	4079	408386	0.2	0.2	0.2	0.2	46.8	1.0	6.1
bench_pushpull_3workers_unix	total	5000	128	4079	1225650	0.2	0.2	0.2	0.2	46.8	1.0	6.3

Key observations:

• Unix domain sockets win on latency — round-trip benchmarks are ~12% faster on Unix (pair_latency avg 74μs vs 83μs; reqrep 73μs vs 90μs) with tighter jitter.
• Bulk 64 KiB transfers strongly favor TCP under backpressure — with ACK-windowed flow control, 65536B pair delivery reaches ~7.1k msgs/s over TCP vs ~1.5k/s over Unix (4.9×): smaller Unix-socket buffers fragment large writes into many more wakeups (avg inter-arrival gap 686μs vs 141μs).
• Small-message floods saturate around 350–500k msgs/s per connection for 512B payloads; PUSH/PULL with 128B frames to a single worker exceeds 1.1M msgs/s on both transports.
• Multi-worker PUSH/PULL divides fairly but adds per-worker overhead — 3 workers aggregate ~496k msgs/s on TCP vs 1.2M msgs/s on Unix.
• Pub/sub fan-out stays imbalanced on TCP — pubsub_2sub delivers 5000/4143 across subscribers before the first completes the run; Unix fans out evenly (5000/5000).
• Backpressure-safe by construction — flood volumes beyond powpow's maxWriteBufferSize cap (default 32MB, tunable via -d:maxWriteBufferSize=N in MB) are handled by ACK windows + watchdog timers instead of silent connection death.

❤ Contributions & Support

• 🐛 Found a bug? Create a new Issue
• 👋 Wanna help? Fork it!

	Switch to Open-Source LLMs via OpenCode GO, choosing from a variety of powerful models such as DeepSeek, Qwen, Kimi, GLM-5, MiniMax, MiMo. 🍕 Use our referral link to get started!

🎩 License

MIT license. Made by Humans from OpenPeeps.
Copyright OpenPeeps & Contributors — All rights reserved.