Never verified
Files copied...never opened. Data transfer interrupted leaves files correctly named and sized with incomplete content. Listings appear intact until someone attempts to use them.
Commonest
There is a difference between having a backup and being prepared for disaster recovery. The first question is "Do we still have our files?" The second question is "How long will it be before we are trading again and what will we lose while waiting?" Many organisations have answered the first question, assuming that it has answered the second.
We are a recovery lab and not a backup service; therefore, we cover the half of the equation that most people never plan for which is what to do if your restore does not work, and how your data will come back from the machine(s) that caused the problem.
Uninspiring; and they convert a vague intent into something that can be tested.
The Recovery Point Objective is the amount of data you can afford to lose in terms of time. For example, nightly backups give you an RPO of no greater than 24 hours. This means that a system failure at 4 pm on a given day would result in loss of all data created since the last nighttime backup. Therefore, whether or not a nightly backup strategy provides adequate protection against losing data depends on whether that loss is acceptable, no matter how reliable the backups are.
The Recovery Time Objective is how long your organisation can afford to be "down". This is the one most organisations have yet to actually calculate. It can take days, using business broadband connections to restore terabytes of data from cloud storage. Your backups may be pristine but the organisation remains shut down.
Both numbers are decision-making issues, not technical realities. Both numbers should be determined by those who carry the costs associated with downtime, not those who manage the storage.
Each of these was brought to us as part of an organisation that felt it had adequate protection.
Files copied...never opened. Data transfer interrupted leaves files correctly named and sized with incomplete content. Listings appear intact until someone attempts to use them.
There is an internal NAS within the same building that can be accessed by the compromised system. A malicious ransomware attack will automatically target mapped drives as well as network shares intentionally. And a single power event can take out both.
Although the job had failed, no alert was seen, and therefore the job hasn't run since. And it is discovered exactly when you need it.
Only documents were backed up; database files and email store files were not; they are locked while the process is active, so the job skipped them silently.
Old advice still right and each part addresses a specific failure mode.
A copy of the data, plus two back up copies. With only two copies, you are one failure away from none. Backup media will fail at the same rate as all other things.
Two separate units to hold the data. Not two separate folders on the same unit. Different pieces of equipment. Different technologies. So if there is a failure mode in either piece of equipment it will not affect both.
Off-site storage can be physical storage of the data outside your normal location. Off-site storage can also be a service that allows for access to the data but does not allow access by using your normal network credentials. Therefore the off-site copy of your data should survive fire, theft and ransom-ware attacks.
Data storage that cannot be changed or removed for a set time. It is the only configuration that will consistently survive an attacker who has access to your admin credentials.
The single most valuable hour of any business reading this.
Find a real file from a month back and restore it. Measure how long it takes, including finding it. Then find something big – a database, email stores, a project directory — and measure as well.
You will usually find out in an hour what nobody wanted to know: Nobody had the password, the restore took longer than expected, a critical drive was no longer backed up after being relocated or the guy who set it up has left. All are inexpensive to find out next Tuesday; all are expensive to find out during an incident.
Image the affected system first: restoring over a system on which an event has occurred will remove all evidence of what happened. Insurers, auditors and the ICO may need to view the system as it was at that point in time when they are investigating or dealing with your claim. In Manchester, business recovery costs start from £500 + VAT once we complete our initial 48-hour diagnostic for free.
Backups are simply copies of data. Disaster recovery is simply planning to restart operations — which covers the copy, but also how long the restore takes, who does it, and what the business does in the meantime.
A recovery point objective is how many hours/minutes/seconds of loss you can afford to take. For example if you back up daily then your RPO is up to 24 hours. A recovery time objective is how long you can afford to be offline. Most businesses do not know their own RTO.
Not on its own. One NAS (network attached storage) device on the same network is simply a second copy of the first device in another box; both copies are exposed to the same power event, the same ransomware, and the same fire or theft. When there are two copies where neither can reach each other it becomes a backup.
Restore from it. Select one of your actual files from last month and do a big file like a database or email store and time how long it takes. This hour will show you missing credentials, what is taking longer than expected and which job(s) have been quietly stopping the coverage of items.
You will have three copies of your data, on two forms of media/systems, with one offsite. Each portion addresses a specific failure, and including one immutably copied version is the sole form that will be able to survive an attacker having access to your administrator credentials.
First, image the affected system. Replacing the original with a restored copy eliminates the state that auditors, insurers or the ICO may need to review — and there is no cost associated with maintaining it. You can always restore after; you cannot un-restore.