Milliseconds and Money: Where WAN Latency Comes From

Latency budgets, queueing math, and the two costs: physics once, queues forever.

Latency bills arrive twice. The speed-of-light part you pay once, in circuit choice. Queues you finance forever, in buffer settings.

The physics bill

Fiber through West Texas beats satellite through a storm. Round trips between a branch and its cloud region cost real milliseconds — the money math for a 40 ms vs 12 ms path on Voicemail, screenshares, ERP rounds. No QoS heroics fixes a geography problem: pick underlays closer to the workloads.

The queueing bill

Scheduler and buffers: shallow queues for voice, paced shapers for bulk, fair queues behind backup storms. Then verify with tests during load, not the lab with the polite graph.

• Measure p95, not averages — averages hide the doctrine
• Mark packets at trust boundaries only
• Re-baseline after every provider maintenance and circuit change

Budget latency like money: allocated, watched, loaned to bulk when the phone lines are off-hours — no obligation on the meeting after.