ZeroMQ Messaging in Python — Core Concepts
ZeroMQ (ØMQ) is a messaging library that embeds directly into your application. Unlike RabbitMQ or Kafka, there is no separate broker process. Your Python programs open ZeroMQ sockets and exchange messages peer-to-peer, which removes an operational dependency and cuts latency.
Socket patterns
ZeroMQ provides several built-in patterns, each suited to a different communication shape.
REQ/REP (Request-Reply). The simplest pattern. A client sends a request, waits for a reply. Think of it as a synchronous function call across a network. Good for RPC-style interactions.
PUB/SUB (Publish-Subscribe). One publisher broadcasts messages; many subscribers receive them. Subscribers can filter by topic prefix so they only get relevant data. This is the pattern stock-feed systems and event buses use.
PUSH/PULL (Pipeline). Work distribution. A ventilator pushes tasks, workers pull them, and a sink collects results. ZeroMQ round-robins tasks across connected workers automatically.
DEALER/ROUTER. Advanced asynchronous patterns. DEALER can send without waiting for a reply; ROUTER tracks which peer sent what. These build non-blocking services and load balancers.
How messages flow
ZeroMQ sockets are not raw TCP sockets. They handle framing, reconnection, and buffering internally.
- Automatic reconnection. If a peer disappears, ZeroMQ retries in the background. When the peer returns, messages resume.
- High-water marks. Each socket has a send and receive buffer limit. When the buffer fills, ZeroMQ either blocks or drops messages depending on the socket type.
- Multi-part messages. A single logical message can contain multiple frames. This is how you attach routing envelopes or separate headers from payloads.
Using pyzmq in Python
The pyzmq package is the official Python binding. A minimal PUB/SUB example:
Publisher:
import zmq
ctx = zmq.Context()
pub = ctx.socket(zmq.PUB)
pub.bind("tcp://*:5555")
pub.send_string("prices AAPL 182.30")Subscriber:
import zmq
ctx = zmq.Context()
sub = ctx.socket(zmq.SUB)
sub.connect("tcp://localhost:5555")
sub.setsockopt_string(zmq.SUBSCRIBE, "prices")
msg = sub.recv_string()The subscriber only receives messages whose topic prefix matches its subscription filter.
Polling and multiplexing
Real services manage multiple sockets. zmq.Poller lets you wait on several sockets at once, similar to select() in network programming. This avoids dedicating a thread per socket.
Common misconception
“ZeroMQ is just raw sockets with a nicer API.” It is more than that. ZeroMQ handles message framing, asynchronous I/O, reconnection, and load balancing internally. Writing equivalent functionality on raw TCP would take thousands of lines.
When to choose ZeroMQ
Pick ZeroMQ when you need low-latency, high-throughput messaging and can handle durability yourself. If you need guaranteed delivery, message persistence, or dead-letter queues out of the box, a broker like RabbitMQ or Kafka is a better fit.
One thing to remember: ZeroMQ gives you powerful messaging patterns — pub/sub, pipelines, request-reply — without running a broker, but you trade built-in durability for raw speed and flexibility.
See Also
- Python Adaptive Learning Systems How Python builds learning apps that adjust to each student like a personal tutor who knows exactly what you need next.
- Python Airflow Learn Airflow as a timetable manager that makes sure data tasks run in the right order every day.
- Python Altair Learn Altair through the idea of drawing charts by describing rules, not by hand-placing every visual element.
- Python Automated Grading How Python grades homework and exams automatically, from simple answer keys to understanding written essays.
- Python Batch Vs Stream Processing Batch processing is like doing laundry once a week; stream processing is like a self-cleaning shirt that cleans itself constantly.