Billing & Charge Model

A budget threshold is a notification rule, not a spending cap. Crossing it emits a message about money already spent. It does not pause running resources, block new ones, or cancel anything, and the figure it fired on lags the usage.

Transfer is metered by direction. Bytes arriving from outside are normally unmetered; bytes leaving toward the internet are priced per gigabyte after a small monthly allowance. So delivery to viewers is the charge, not the upload of masters.

Providers meter usage along several independent dimensions: running time, requests served, gigabyte-months stored, gigabytes moved, and provisioned capacity units. One service can charge on all of them at once, so a large bill line exists with no hourly rate anywhere.

Storage and reservations outlive the machine. A block volume that was not marked to be deleted with it, the snapshots and images built from it, and a reserved public address all keep billing. Find them by listing resources attached to nothing, not by reading the machine's record.

Charge records carry the labels present when usage was metered, so a late label never repairs the days already billed. Short-lived resources are born and destroyed between sweeps and are never attributable at all, while creation is a single point that catches everything once.

A usage commitment fixes a quantity of a specific resource shape per hour, so only that shape earns the discount. A spend commitment fixes an hourly amount of money and applies across whatever usage qualifies, trading depth of discount for flexibility.

Traffic is metered by the boundary it crosses: out to the internet, between regions, between zones inside one region, and through a managed gateway it is forced through. The in-region crossing is missed because nothing in the design looks remote.

The full declared level, for the whole hour. Provisioned pricing meters the capacity you asked for and held, not the work done, so an idle hour costs the same as a busy one and utilisation becomes the number to watch.

Anything metered by allocation rather than by use. Machine hours, managed-store instance hours, provisioned throughput floors, allocated block storage and per-hour licences all bill for existing. Only consumption-metered charges - requests served and bytes moved - fall at zero traffic, and they were the small part.

Unit cost: spend attributed to the service divided by a unit of work the business already counts, such as completed orders. A total confounds growth with regression, while a per-unit figure stays comparable as volume changes and survives the growth argument.

Coverage is the share of eligible usage billed at a committed rate; utilisation is the share of the commitment actually consumed. High coverage with low utilisation means the commitment was sized above the workload's floor, and the unconsumed part is cash spent on nothing.

Every charge belongs to exactly one account by construction, so account-level attribution has no unlabelled remainder and needs no convention to be agreed first. The price is granularity: it resolves spend to a whole account, not to the teams sharing one.

Cost data is metered continuously but published hours later, and early figures are restated until the period closes. So cost data confirms a spend change rather than detecting one. Detect on the usage counters your workload already produces.

Three dials set discount depth: the length of the term, how much of it is paid before it is consumed, and how narrowly the commitment is scoped. Every extra point of discount is bought with certainty for the provider and optionality for you.

Rounding increments and minimum billable amounts. A meter charges in whole units of its increment and never below its minimum, so very short or very small units pay for time and volume they never consumed, sometimes several times over.

Because the charge is a stock, not a rate: every run adds stored bytes and nothing removes them. A retention rule bounds it, at roughly the nightly change volume times the days retained. Without one, the stock rises for as long as the job runs.

You cannot label a genuinely shared cost into one team's bucket, so you choose a posture instead: leave the pool visible and unallocated, or divide it by a driver everyone agreed to in advance. Any split is a convention, not a measurement.

A flat threshold treats every day as identical, so a recurring weekly peak always looks anomalous. A useful baseline compares like with like — this Monday against recent Mondays — using a robust centre and a minimum absolute money delta before anything is reported.

Every inter-tier call now crosses a metered zone boundary, so the charge scales with calls multiplied by payload, not with machine count. Measure gigabytes crossing a boundary per request before changing any part of the design.

Split the service line into its charge units and multiply each quantity by its own rate. Rank the dimensions, then ask which design decision produces the top quantity — call count, retained volume, held capacity or elapsed time — and attack that one.

Join three sources: the platform inventory of everything that exists, the billing records that price it, and last-use evidence appropriate to each resource kind. Rank the unreferenced rows by monthly charge, attach an owner and the evidence, and quarantine before deleting.

Commit only the part of the baseline the redesign will not touch, on the shortest and most flexible instrument that still pays. Flat spend proves the amount is stable; it says nothing about the shape, and a stranded term keeps billing against a workload that no longer exists.

Most mid-period forecasts are run-rate extrapolation: spend so far divided by days elapsed, times days in the period. That breaks when a one-off charge gets multiplied across the period, a monthly-cadence charge is treated as daily, or stored volume is still growing.

On a graduated schedule only the units above the threshold get the cheaper rate; earlier units keep the rate of the tier they fell in. Nothing is promised — the rate falls automatically with consumption inside a billing period.

Because meters attach to boundaries and to components, not to bytes. One gigabyte can be charged for crossing a zone boundary, again for being processed by a gateway in its path, and again at the internet boundary on the way out.

The pilot sat inside a free allowance on that dimension. An allowance hides the rate rather than lowering it, so the first usage above it is charged at full price and the bill steps from zero to a real number.

Labelling the resource is not the same as labelling the charge. Charges generated by resources a managed service creates for you may not inherit the parent's label, a label key often has to be enabled as a reporting dimension first, key drift splits one owner across buckets, and some charges have no resource at all.

Utilisation alone cannot tell them apart. A low ratio only proves the floor exceeds observed demand; deliberate headroom has a named reason and an event that consumes it. Get consumption across a full business cycle plus an owner's stated reason, then step the floor down rather than cutting it.