Pub/Sub Best Practices: Publishing

Chloe, whose affiliation is not established by the source, presents six publishing practices centered on explicit throughput, latency, ordering, and retry tradeoffs.

Key points

• Batching exchanges latency for throughput and request efficiency: Message size, publisher count, flow control, cost, and latency objectives should determine settings rather than one universal batch size ([00:00:00]-[00:01:03]).
• Ordering keys narrow the safe publishing pattern: The episode recommends one-message batches and waiting for server acknowledgement before publishing the next message for that ordered stream ([00:01:03]-[00:02:06]).
• Its ordered-delivery recommendation is regional: It says ordering is guaranteed only within a region and recommends a regional endpoint so messages using ordering keys enter the same region ([00:01:03]-[00:02:06]).
• Publishers have client-side resource limits: Bandwidth, network latency, request rate, CPU, memory, and threads constrain throughput, while flow control can bound pressure ([00:02:06]-[00:03:10]).
• Retry settings need measurement: The episode identifies initial timeout, retry multiplier, and total timeout as tunable rather than universally safe defaults ([00:02:06]-[00:03:10]).

Current-use note: Treat the regional ordering, client behavior, and retry specifics as 2023 guidance. Verify current documentation and benchmark the actual library and workload; retries can still create duplicate application effects that require idempotent handling.

Full video: https://www.youtube.com/watch?v=Oo5PXA4idI8(opens in a new tab)