For the person doing the evaluating
Eight steps, in order, each one ending in something you can show a colleague. Our setup guide tells you how to install a gateway. This page is the other job — deciding whether you should, and having an answer you can defend.
Put in what you store, what you read back and how long you keep it. The calculator prices XNS and five other vendors on the same inputs, and it reads competitor rates from a published file rather than from copy we wrote.
You end with: a monthly figure you can put in a budget line. Cost calculator → · Rates → · Why egress dominates →
Storage provider sets are composable by geography and jurisdiction. Work out what your policy actually requires — a country, a legal jurisdiction, a distance from a primary site — before you look at whether we can meet it, because the requirement is usually vaguer than people expect.
You end with: a written residency requirement. How placement works →
Not “is it S3-compatible” — every vendor says yes. Check the specific operations your software calls. We publish the conformance run operation by operation, including what fails and why, with the raw counts and the date of the run.
You end with: a list of anything your stack needs that we do not do. Compatibility matrix → · Tool-by-tool setup →
Encryption, key custody, identity, the audit surface, and the line between what you operate and what we do. Also the durability model with its assumptions written out — including the failure mode it does not cover. Bring your reviewer’s specific question; most of them are already answered on the page.
You end with: answers for your security questionnaire. Trust & security → · Durability model → · What runs, and how you would leave →
About ten minutes. Use test data, not production data — the point of this step is to confirm the thing exists and behaves, not to migrate anything. Keep the gateway; you need it for steps 6 and 7.
You end with: a working S3 endpoint on your own host. Get started → · Full setup guide →
A representative slice, not a tidy one — include your largest objects, your deepest prefixes and your worst filenames. Use the tool you would use in production, verify checksums after the copy, and record how long it took. Time it yourself rather than taking a number off our page.
You end with: a measured ingest rate on your own link. Setup for rclone, restic, Kopia and thirty more →
This is the step most evaluations skip and the only one that tells you whether the plan works. Pull the sample back, verify it against the source, and record the wall time. Restores are not metered here, so a drill costs you nothing but the hours — which is the argument for drilling monthly once you are live, not just once during evaluation.
You end with: a restore time you measured. What is and is not supported for backup →
Bring what your process requires and we will tell you plainly what we have, what is in progress, and what we do not offer. An honest no is faster for both of us than a slow maybe.
You end with: a decision you can defend. Talk to us →
Most storage evaluations take six weeks because the decision is hard to undo. Getting data out of a metered platform costs real money at exactly the moment you have decided to leave, so the evaluation has to be right the first time.
Step 7 is the same operation as leaving. If a restore drill works and costs nothing, then so does the exit — at any size, at any time, with any S3 tool, without asking us. That is what shortens the evaluation: not that the decision is easier, but that it is cheaper to be wrong.
The exit path is written out with commands on the production readiness page →
Size the bill before you install anything. It takes two minutes and it decides whether the other seven steps are worth your time.
Explore the platform
Everything you need to go from evaluation to production.