Skip down to the research record

Immutable backup costs: 2 rates to compare

Quota Forecast Editorial Team · Published · 10 min read

Immutable backup costs combine storage charges for the terabytes you keep and request charges for the calls your backup software makes.

Key takeaways

  • Measure stored data, including retained copies.
  • Count uploads, lock writes, and renewals separately.
  • Compare current storage and request rates using the same protection window.

Both rates are published, so you can compare them line by line once you know how much data you keep and how many calls your backups make.

This post follows one example case from start to finish. A revenue team keeps 1 TB of forecast files in an immutable backup. Here, "immutable" means the stored copies cannot be changed or deleted until a retention period configured in the storage service has passed. The backup job writes the files and sets or renews a lock on them on a regular schedule. Each section below returns to that team and its 1 TB.

Most of the figures come from two Backblaze blog posts. The AWS request price appears inside one of those posts, so the link points there. Prices change, and the figures here are the ones on those pages when we wrote this post. Check current pricing before you act on any number.

Two rates, two different jobs

Your backup bill has two moving parts. The first is storage. You pay a rate for the data you keep, and the charge depends on how many terabytes sit in the backup and how long they stay there. The second is requests. Each time the backup software writes a file, or sets or renews a lock on a file, it makes a call to the storage service. Some providers charge for those calls and some do not.

The two rates behave differently. Storage is a steady charge that repeats every month for as long as the data stays. Request charges depend on how your backup tool behaves. A tool that writes more often, or renews protection more often, makes more calls and pays more each month, even when the amount of stored data stays the same.

That difference is why the rates need comparing together. A storage price for each terabyte is easy to find and easy to assume is the whole bill. A request price often sits elsewhere on the same page, is measured in a different unit, and grows with activity rather than with data size. If you compare only the storage line, you may miss the line that changes most with how the backup job runs.

Storage: what one terabyte costs each month

Start with the storage rate. Backblaze quotes $6/TB/month, or $6,000 a month for one petabyte. The source is Backblaze, "Storage Tech of the Future: Ceramics, DNA, and More".

At that rate, the team's 1 TB storage line is $6 a month. The petabyte figure in the same sentence agrees with the rate for each terabyte, so the storage line rises in a straight line. Each added terabyte adds the same $6 a month, and a team that keeps several times as much data pays several times that line.

Three things change the storage line in practice, and none of them requires a new price:

  • How much data you keep. The storage line counts what sits in the backup, not what the forecast model produced. If the backup holds more than one copy of a forecast set, each copy counts toward the terabytes you pay for.
  • How long you keep it. A terabyte held for one month costs the rate once. The same terabyte held for longer costs again for each extra month. Retention periods in your policy therefore show up in the bill.
  • What unit the invoice uses. The rate is for each terabyte each month. Check whether your invoice uses decimal or binary terabytes, and whether it rounds partial amounts up or down.

Forecast archives also grow. Each new model version or export adds to the stored size, so the total often rises even when the team does not change its process. Check the stored size every month instead of relying on the figure you measured at setup. A storage line that was right in the spring can be low by the autumn. For the files behind that growth, see weekly pipeline export checks.

For planning, use the stored size at the end of the retention period rather than the size on the day the backup starts. Multiply that figure by the rate to estimate the storage line at that size. Budgeting from the starting figure understates the bill in later months whenever the stored size grows.

Take the AWS storage rate from AWS's own pricing before you finish the comparison.

Requests: what write charges add up to

A PUT request is a write call. Storing a file, overwriting it, and in many setups setting a lock on it are all writes to the storage service. Providers usually quote the price for a block of calls, and the block size matters for the arithmetic.

The Backblaze post on object lock reports an AWS S3 price of $0.005 for 1,000 standard PUT requests. Applying that price to lock-related writes and renewals is the author's reading of the source and needs checking against it. The source is Backblaze, "Object Lock Mistakes That Put Immutable Backups at Risk".

Read the unit carefully. The price is $0.005 for 1,000 requests, not $0.005 for each request. Divide the call count by 1,000, then multiply by $0.005. A single call costs a tiny fraction of a cent, and a reader who skims the number can misjudge the total by a factor of 1,000 in either direction. Write the unit down next to the number.

For the example team on AWS, the request line is the monthly call count divided by 1,000, times $0.005. The team's call count is not known here, so the formula is given instead of a made-up total. The count can come from the backup tool's logs or the provider's request metrics.

Check whether the quoted PUT rate applies to each operation before including it in the calculation. A team that counts only new file uploads may understate the request line. Backup tools do not all work the same way, so read your tool's documentation to see which operations send write calls.

Each 1,000 calls adds $0.005 to the bill. That step is small and easy to miss on one invoice, but it repeats every time the backup job runs. The request line is a running total of activity, so its size depends on how often the job runs and how many lock operations each run makes.

Think of the request line as a monthly meter that runs whether or not anyone opens the forecast files. A job on a fixed schedule produces roughly the same calls each month, while a job that runs on demand follows the team's activity. Compare several months before you decide the request line is stable.

Object lock renewals and the bill

Renewals are where the two rates meet. Backblaze describes how its service handles them: "Backblaze B2 doesn’t charge for PUT requests, so lock writes and renewals are included with your storage." It then draws the consequence: "In practice, this means shorter immutability windows don’t carry a cost penalty on B2, and there’s no incentive to minimize lock renewals to reduce your API bill."

Backblaze is describing its own service, so confirm that position against current pricing terms before you rely on it.

For the example team, this changes the decision. On B2, the request line for lock writes and renewals is $0 according to that post, so the $6 storage line for 1 TB covers this part of the bill. Choosing a longer or shorter immutability window does not change that arithmetic. The choice becomes one of policy: how long the forecasts must stay protected, and how far back the team needs to recover a file.

Recovery planning deserves the same attention as cost. A window that is too short can leave the team unable to restore an older forecast version it needs to explain a past number. The right window is the one the team's own records requirements call for. A cost review should fit the budget to that window, not shrink the window to fit the budget. See also backup recovery checks before quarter-end.

On AWS, confirm whether lock writes and renewals fall on the quoted PUT line before pricing them. Letting the invoice drive the window carries a real risk. A shorter window or fewer renewals may lower the request line, but it may also leave files protected for less time than the policy requires.

A practical brief for a revenue leader fits on one page with four items. State the protection window the policy requires. Show the B2 line: the storage rate applied to the terabytes kept, plus $0 for lock calls according to the Backblaze post. Show the AWS line at the team's own call count. Then state the trade-off in one sentence.

Putting the example together

The monthly model has two lines, one for each rate:

  • Storage line = terabytes kept × monthly storage rate for each terabyte.
  • Request line (AWS) = monthly calls ÷ 1,000 × $0.005.

The table puts the quoted rates side by side:

LineQuoted rate
Storage on Backblaze B2$6/TB/month
PUT requests on AWS S3$0.005 for 1,000 requests

On the Backblaze side, the example's storage and lock calls cost $6 a month. Adding 1 TB raises the storage line by $6. On the AWS side, adding calls raises the request line by $0.005 for every 1,000 calls. Data growth moves the bill in dollars for each terabyte, while call growth moves it in fractions of a cent for each 1,000 calls.

Whether storage or requests dominate depends on the team's actual call count and its AWS storage rate. A storage rate from one provider and a request rate from another cannot be compared until both sit in the same monthly formula for the same data.

Use the formula to answer three questions. What does the storage line cost at the terabytes you keep? What does the request line cost at the calls you make? Which provider's total is lower once both lines are added for the same protection window? If you cannot answer the third question, the first two are not yet complete.

Illustrative example: storage arithmetic

For an illustrative backup holding 1,000 TB, apply the quoted storage rate to each terabyte. The storage line is $6,000 a month. Lock calls remain separate from this storage calculation. This is an illustrative storage line, not a complete bill.

What to check before you compare

Work through these checks before you compare invoices or sign a renewal:

  1. Terabytes kept. Pull the stored size from the backup tool's report for the month. Include every retained copy and every retention period your policy requires.
  2. Calls made. Count the month's calls from the backup logs or the provider's request metrics. Count uploads, lock writes, and renewals separately so you can see which one drives the request line.
  3. Units. Confirm that each storage rate covers a terabyte for a month and each request rate covers 1,000 requests. Do the arithmetic in that unit and nowhere else.
  4. Current prices. Read each provider's pricing page on the day you compare.
  5. Other charges. Ask each provider whether any charge applies beyond storage and requests for your setup. Keep the reply on file with the comparison.
  6. Retention first. Set the protection window from your policy, then price it. Do not shorten the window just to reduce the request line.

Most of these checks take a few minutes once the reports are open. The slow part is usually the call count, because it lives in logs that many teams never read. Start there.

Limits of this comparison

This post relies on two vendor blog posts. Backblaze writes about its own service and about AWS prices, so treat its statements as the vendor's position and confirm them against current pricing terms.

The example is one team with 1 TB. Your data size, copy count, retention period, and call count will differ, and the arithmetic scales to them. The formulas are the part to reuse. The numbers are the part to replace with your own.

The comparison also covers only the two rates in the title. Recovery speed, support terms, and other charges on a provider's price list can change the total, and each needs its own check against that provider's current terms.

Nothing in this post is a contract quote. Use it to decide which lines to ask about and which numbers to request. When you take the comparison to finance or to a provider, bring the formula, the unit for each rate, and the call count from your own logs. Those three things make the conversation shorter and the result easier to check.

Today, open the backup logs and record the month's uploads, lock writes, and renewals separately beside the stored size.

Questions

Is the storage rate enough to compare backup bills?

No: compare storage and request charges for the same data and protection window, then check other applicable charges.

Where can I find the monthly call count?

Use the backup tool's logs or the provider's request metrics, checking which operations the count includes.

Sources