Area 07 · Operations and infrastructure

Backup and restore

The console reported that last night's job finished. That fact is worth almost nothing on its own. The question worth answering is whether your finance team can open last week's ledger on Monday morning once ransomware has swept through the file server. So we begin at the far end, with the outage each application can absorb, and let that dictate everything upstream of it. Then the whole thing is rehearsed against a clock, at an interval you choose. Drills included, none of it requires anyone to travel.

3-2-1
copies on two kinds of media, one held elsewhere
Drill
a rehearsed recovery, measured in minutes
EU
the off-site copy remains within Europe
Plan
agreed while nothing is on fire

Set two numbers first

Every system needs an RTO, meaning the longest outage it can absorb, and an RPO, meaning how much recent effort you would be willing to key in again. What follows are plausible figures for a firm of thirty or forty people.

Accounting and ERP

Four hours down, one hour of data. Copying a Sage, Cegid or EBP database once at midnight leaves too much exposure on the day a VAT return goes out. Transaction logs want capturing repeatedly between nine and six.

Shared folders and SharePoint

A day either way. One pass taken after the office empties covers what most companies genuinely require.

Microsoft 365 mail

Four hours down, one hour of data. Keeping the platform running is Microsoft's commitment. A mailbox stripped bare by somebody working out their notice is not, so mailboxes want a backup of their own.

Online shop

An hour down, an hour of data. Orders from Amazon.fr and card authorisations do not pause politely while you rebuild from an archive, which leaves a standby environment kept warm as the only honest answer.

Laptops

A day either way. Realistic only where work is saved into OneDrive or SharePoint rather than a desktop folder on one particular machine.

A second folder on the very same server protects nothing at all. One controller failure, one encryption run, one careless drag of the mouse, and original and duplicate depart together. Two copies on unlike media, plus a third somewhere else entirely, ideally rented from a European host: that is the floor, not the target.

How the work is done

Reaching your servers and the Microsoft 365 portal happens over encrypted remote sessions throughout. Building the thing takes days. Stage four is what turns a configuration into a guarantee, and it is routinely left out.

01

Locate the data

Important files rarely sit where an organisation chart suggests they will. Recurring surprises: a price list living on one sales laptop, or executed contracts that exist nowhere except as mail attachments.

02

Agree the targets

Each application receives a tolerable outage and a tolerable loss, decided together with you. Budget follows from those numbers; it is not allowed to set them.

03

Build it

Timings, retention, encryption, the off-site leg. Alerting covers failures and, just as importantly, reports that never turn up at all.

04

Prove it

On the agreed cycle, chosen data is rebuilt inside a sandbox and the clock is read. What comes out is a dated note you can put in front of an insurer or an auditor.

Questions and answers

Put one question to whoever owns them: what was the last thing they brought back? Typical discoveries include a job erroring since spring, a protected folder that stopped being used two reorganisations ago, and an encryption passphrase that walked out of the door with a leaver. Our review ends in a yes or a no, not in a set of suggestions.

They can, provided the credentials that open your network do not also open them. Current criminal practice is to find the repository and empty it first, encrypting afterwards, specifically so that paying becomes the only remaining option. What defeats that is one copy behind a sign-in of its own, frozen against modification for a fixed number of days.

Thirty dailies, thirteen weeklies and twelve monthlies make a reasonable opening position. Accounting records run far longer, since French commercial law expects several years of them, and health records answer to separate rules again. Those durations belong to your accountant or your lawyer; we turn whatever they specify into settings.

The regulation asks, in so many words, that you be able to bring personal data back within an appropriate time. Organisations inside the NIS2 perimeter are expected to manage continuity and to show their working. Dated recovery notes are exactly that evidence, whether the question arrives from an auditor or from the French supervisory authority. Whether NIS2 catches your organisation is a legal matter, and not ours to rule on.

Either a colleague on your side, talked through it stage by stage, or the engineer you already call locally. Hardware is neither shipped nor fitted by us. What we contribute is the diagnosis, the exact bay number, and every task achievable across the wire: array rebuild, data recovery, and the consistency checks that follow.

We find out whether your backups really restore

Your present configuration gets examined remotely, a genuine recovery is attempted from it, and you hear a straight answer about whether it would carry the company through a bad day.

When we are around
Weekdays, 8:00 to 18:00 CET; answers land inside one working day
Talking it through
A call on Teams or Google Meet, whenever writing is not enough

We set strictly necessary cookies only: they keep the site running and remember the city you chose. Nothing here is used for advertising or tracking. More in our privacy policy.