Considerations of a Distributed Messaging Queue’s Design

Message order needs a scope and a definition. Before selecting an algorithm, decide whether the application needs one producer’s sequence, a broker partition’s accepted order, causal dependencies, or an approximate event-time view.

The design must then preserve that contract through concurrent producers, delayed messages, consumer failures, and retries. Each choice affects buffering and processing capacity.

Ordering of messages

A chat conversation may need an agreed order within that conversation, while unrelated conversations can proceed independently. An email workflow may tolerate reordering for some messages but still require a verification step before a dependent notification. Define the business dependency; the application category alone does not settle ordering.

We categorize message ordering into two types:

• Best-effort ordering

• Strict ordering

Best-effort ordering

A best-effort contract tries to preserve a useful order without guaranteeing it in every case. In the simplified example below, the broker accepts arrival order and does not reconstruct the producer’s send order.