Affordable Archive Storage for a Photography Studio

"All my photos" is not a thing you can make decisions about. Split the archive into four tiers and the decisions get easy, because the tiers differ by a factor of a hundred in size and by a factor of a hundred in how often anyone needs them.

The last row is the one to notice. The catalog and the edit history are the cheapest thing in the archive and the piece most studios lose. RAW files survive because they are obviously valuable and they get copied. The catalog lives on the working machine, gets replaced when the machine gets replaced, and takes the ratings, the crops, the colour work and the collection structure with it.

Losing it does not lose the photographs. It loses the decisions. A 2017 wedding whose catalog is gone is a folder of eight thousand unranked frames, and the four hundred you chose are no longer distinguishable from the rest. Back up the catalog with the job, not with the laptop. It is the highest return on effort in this guide.

Culling RAW files is the one decision on this page that cannot be undone. Every other mistake here is recoverable by spending money. This one is not, so it deserves a rule rather than a slogan.

The rule, in three parts:

• The selects and the delivered masters are kept, permanently. These are the frames you chose, the frames you finished, and the files the client owns. They are a small fraction of the total volume and they carry nearly all of the value. There is no volume argument for deleting them.
• The unculled take is a business decision, not a storage decision. It is where the terabytes live, and how long you hold it should follow your genre, your contract and your licence terms. Commercial work under a live licence keeps its originals through the licence window. A wedding keeps its originals while album and reprint requests are still plausible. A conference weekend that produced twelve thousand frames nobody will ever reopen is the honest candidate for a shorter period.
• Cull once, at delivery, then never again. Retroactive culling is where errors happen, because you are years past remembering which frame was the one. Rejecting the obvious failures while the shoot is fresh removes far more volume than a nervous pass in 2031 ever will.

Two things that are not culling and should be done first, because they free space at zero risk: delete the duplicate copies of the same job that past migrations scattered across drives, and delete anything you can regenerate, such as previews, caches and export folders whose contents are already in the delivered set. On a decade-old shelf that pass often recovers a meaningful share of the space before a single original is touched. And if any of this bumps into a promise you made a client, the promise wins.

The widely used shape for a backup is called 3-2-1, and the Canadian Centre for Cyber Security states it plainly: "Adopt the 3-2-1 rule for data storage and have 3 copies of your information (1 original and 2 backups), saved on 2 different media types, with 1 copy kept off site."

Translated into a studio:

• Three copies. The working copy on the machine or the production volume, plus two more. A copy that exists only on the drive you edit from is not a copy of anything.
• Two media types. Two drives from the same box, bought together and spun up together, fail in a correlated way. Different technology, or different hardware and a cloud target, breaks that correlation.
• One copy off site. This is the part people skip, and it is the part that covers the events that take the whole studio at once: fire, flood, theft, a burst pipe over the equipment shelf.

3-2-1 is not redundancy for its own sake. It is a way of making sure no single event can reach every copy. Ask that of your own setup: what one thing, happening in one room, takes out everything you have.

There are three ways a small studio can hold an archive. Each is genuinely the right answer for someone, and each fails in a specific way.

Local drives and a network volume. Cheapest per terabyte by a wide margin, bought outright, entirely under your control, and the option that fails silently. A drive unread for four years can be dead without announcing it, and a network volume with a redundant disk protects you against a disk dying, not against a deletion, a bad copy or an encryption event that reaches every disk in the box. Rotate: keep one drive that lives away from the studio and swap it on a schedule you actually hold. Consumer drives are not archival media, and an untested backup is not a backup, it is an assumption with a label on it.

Consumer cloud sync. Convenient, already installed, and not designed as an archive. Sync means changes propagate, which is the opposite of what an archive wants: a deletion or a bad overwrite travels to the cloud copy at the speed of your connection. Sync services are also tier products. You buy a fixed size and you either fit in it or you buy the next one up, and the next one up is often a team plan with a seat minimum you do not need.

Cold object storage. Billed per gigabyte actually stored, with no tier to buy, so the bill follows the archive instead of the archive having to fit a plan. Downloads are billed by the gigabyte as well, which is the trade for a low storage rate and the reason this is a home for the copy nobody opens. The honest trade is the retrieval experience, not the price: this is not a folder you browse. Getting a job back is a task, run with a tool or a script, taking as long as your connection takes. It is the right home for the copy you hope never to open, and the wrong home for anything you touch weekly.

Read that table carefully, because the obvious conclusion is the wrong one. At two terabytes the consumer plans land in the same neighbourhood per terabyte, and the thing that separates them is not the price. Those prices were served in Canadian dollars and no conversion has been applied, and the per-terabyte-per-year column is the published monthly price multiplied by twelve and divided by the plan storage.

The difference appears when the archive outgrows a consumer tier, which a decade of RAW does. Take a shelf holding six terabytes. Pay-as-you-go cold storage bills that as six terabytes and nothing more, growing a gigabyte at a time rather than a tier at a time. Neither consumer plan above holds six terabytes. Dropbox's next steps are Standard, which "Starts at 3 TB for the team" at CA$21 per user per month, and Advanced, which "Starts at 15 TB for the team" at CA$33 per user per month and requires three or more users. Three seats you do not need is CA$99 per month for storage you were buying by yourself.

So the rule is not that one option is cheapest. It is that a fixed-tier product prices an archive by the tier while a pay-as-you-go product prices it by the gigabyte, and an archive only ever grows. That is why the off-site copy in 3-2-1 usually belongs on pay-as-you-go cold object storage rather than on a sync subscription: the archive keeps growing without ever forcing an upgrade nobody asked for.

An archive that depends on a subscription staying paid has an expiry date, and the providers say so in their own documentation.

Google states, on its own support page about storage limits: "If you're over your storage quota for 2 years, we might delete your content across Gmail, Drive, and Photos." It is a "might", not a guarantee, and worth quoting exactly for that reason. It is also the published policy, the free tier it drops you to is 15 GB shared across three products, and two years is not long in the life of a lapsed card.

Dropbox is similar in shape. Cancelling converts the account to Basic, which holds 2 GB, and Dropbox's own downgrade page says that if you exceed that quota you "may not be able to sync, upload, share, move, or preview files and folders" and that "If your account remains over the limit, Dropbox may delete files you own to bring your account below the storage limit."

None of this is a scandal. It is what a storage subscription is. The point is structural: if the archive lives entirely inside a paid account, a dead card or a bounced renewal is a data-loss event with a countdown attached. That is the argument for keeping at least one copy on hardware you own outright, and for the dull administrative habit of keeping the billing card current on whichever account holds the off-site copy.

Restore one job a year, at random, and open it. That is the whole practice, and it is the difference between an archive and a hope.

Do it properly, which takes about an hour:

• Pick a job you did not choose deliberately. Roll a year, take the third folder. Testing the job you remember copying correctly tests nothing.
• Restore it from the off-site copy, not the local one. The local copy is the one you already trust.
• Open the actual files. Not the folder listing, the files. A directory of the right size full of unreadable data looks perfect from the outside.
• Open the catalog too, and confirm the ratings and the edits came back. This is where most restores are found to be incomplete.
• Write down the date you did it and what failed. Something usually does the first time, and finding it in a drill costs an hour instead of a client.

Cost is not a reason to skip this. One job is a small fraction of an archive, and on pay-as-you-go cold object storage you pay for the gigabytes you actually pull back and nothing else, so a single yearly restore is a rounding error against the storage line. Read your provider's download terms once, put the drill in the calendar, and stop thinking about it. The drill is cheap. Not doing it is what is expensive.

Briefly and plainly, because overstating this would undo the rest of the page.

clientgallery.io is a delivery system, not an archive. It does not warehouse RAW files. What it holds is the proof layer: photos are compressed in the browser to proof JPEGs of roughly 0.6 MB each and hosted for as long as the account exists, on one plan at $10 a month or $100 a year, with no storage tiers to climb. Fair use is 1,000 photos per gallery and 250 GB of proofs per paid account; a free account holds 5 GB.

Full-resolution delivery works as a handoff rather than as storage. You upload a full-resolution ZIP for a gallery, it stays live for 3 days per handoff, and you can re-upload it any time the client needs it again, with 100 GB of standing ZIPs per account. That design is deliberate. Keeping master files out of the platform is what keeps the price flat and predictable.

The two systems do different jobs and a studio wants both. The gallery is where the client browses, chooses and downloads. The archive is where the negatives live, on your own hardware plus one off-site copy, priced by the terabyte and tested once a year. Anyone who tells you a delivery platform is also your archive is selling you one of those two things and charging you for the other.

• Canadian Centre for Cyber Security, Tips for backing up your information (ITSAP.40.002). cyber.gc.ca/en/guidance/tips-backing-your-information-itsap40002. Fetched 9 September 2026.
• Google Drive Help, storage quota and what happens when you exceed it. support.google.com/drive/answer/6374270. Fetched 9 September 2026.
• Google One plans and prices. one.google.com/about/plans. Fetched 9 September 2026. Prices were served to this fetch in Canadian dollars.
• Dropbox plans and prices. dropbox.com/plans. Fetched 9 September 2026. Prices were served to this fetch in Canadian dollars.
• Dropbox Help, downgrade Dropbox individual plans. help.dropbox.com/plans/downgrade-dropbox-individual-plans. Fetched 9 September 2026.

Product figures for clientgallery.io are the current published plan terms. Per-terabyte annual figures are arithmetic on the published monthly prices above, with no currency conversion applied.