Skip to content

Size your speed tier

Speed tiers are sold by throughput, not by traffic, so the question is not “how many gigabytes will I move” but “how much of that do I need to move at once”. Measure a real batch before you pick.

This page lives here rather than on the product page on purpose: it is a lab procedure, and a lab procedure between the price table and the setup steps costs you the decision it was meant to help with.

Use the actual destinations, route, protocol, session mode, and client you plan to run in production. A small sample against easy targets cannot size a workload with different destinations, locations, or session behaviour.

Record, for both successful and failed attempts:

MeasureWhat to recordWhy it matters
Batch shapePlanned request count, exact target set, requested route, protocol, session mode, test window, total elapsed timeA sample that differs from production in destination or session behaviour does not size production
Response sizeTransferred request and response bytes per attempt, split by successful, challenged, and failedPer-GB cost follows transferred bytes; a fixed tier has to carry the measured traffic inside the window
Success and retry rateSuccessful responses, target rejections, timeouts, other failures, total attempts, retries under a declared policyRetries consume both time and transfer — compare cost per useful result, not per initial request
ConcurrencyPeak simultaneous requests, and how long that peak is sustainedThe speed tier is a ceiling on the peak, not on the average
ScheduleWhether traffic is continuous, bursty, or scheduled, and the hours it must complete inA burst that must finish in one hour needs a bigger tier than the same bytes spread over a day
  1. Take the transferred bytes for one full representative cycle, including retries.
  2. Divide by the seconds the cycle is allowed to take. That is your required average throughput.
  3. Multiply by your peak-to-average ratio — the peak is what the tier has to absorb.
  4. Convert to megabits per second: bytes per second × 8 ÷ 1,000,000.
  5. Choose the next tier up. Tiers are a ceiling; sizing to the exact number leaves nothing for variance.
  6. Re-measure after a week of production traffic and adjust.

Effective throughput still depends on destination behaviour, concurrency limits, and retries. The tier bounds what the network will carry for you; it does not promise what a destination will serve.

At the $1.64/GB entry rate, unlimited becomes cheaper somewhere past roughly 350 GB a month — the cheapest monthly unlimited tier is 100 Mbps at $574.18. On the volume rates (about $0.84/GB at 500 GB) the crossover moves out to roughly 680 GB.

Both numbers assume the traffic fits inside the tier you would buy. If the same bytes have to move in a two-hour window rather than across the month, the tier — not the total — decides the price.