Picture a hospital’s patient portal going dark on a Monday morning. Appointment slots, teleconsultation links and the page where families upload paperwork all vanish at once. Nobody is hurt by an outage like that, but the administrative team feels it within minutes. The discipline that decides how fast such a service comes back, and who gets to say “switch now”, is called DisasterRecoveryOps. Jargon stays light here, and every example is administrative rather than clinical.
What DisasterRecoveryOps actually means
Strip away the vocabulary and DisasterRecoveryOps is a habit: rehearsing recovery week after week instead of filing a binder and forgetting it. xOps School’s reference project, the Ops Atlas, calls this an uncommon label for IT disaster recovery treated as ongoing day-to-day work. It admits the term overlaps with others, so colleagues may use different words for the same thing. Readers who want depth can open the Ops Atlas page on DisasterRecoveryOps; only a few of its ideas appear below.
The practice itself is old. The Atlas traces written guidance back to a 1981 US federal standard, later replaced by NIST’s contingency planning publication in 2002 and revised in 2010. So nothing here is a fad; what changes is how often teams rehearse.
Two numbers every team should know
Two abbreviations come up in every conversation about recovery. RTO is the recovery time objective: how long the service may stay down. RPO is the recovery point objective: how far back in time you can afford to land when the service returns. Notice the second one is a moment, not a quantity. If your portal comes back showing the state of 9:55 when it failed at 10:00, your recovery point sits five minutes behind the crash.
Think of a clinic’s front desk using a paper diary during a power cut. The diary is perfectly usable, but any booking made in the last few minutes before the cut was only on the screen. That gap is the recovery point, and the minutes before somebody reaches for the diary are part of the recovery time.
The clock starts before anyone touches a server
This is the idea that surprises most newcomers: according to the Atlas, the stopwatch also runs while people deliberate, not merely while machines move. Someone has to notice the trouble, confirm it is not a blip, and then choose to invoke the recovery plan. Every one of those minutes counts against your target.
The Atlas offers a worked drill to make this concrete. In its example, people took 31 minutes (six to detect, 25 to decide) while the systems needed 33 more. That adds up to 64 minutes against a 60-minute objective, a miss by four minutes. The surprising bit is that the machines were not slow at all. The hesitation in the room was the biggest slice.
Most teams measure only the technical half, because that is the half with graphs. If you recognise your own organisation in that, you are in good company. A hospital administration office, for example, may time how long a backup system takes to start but never time how long it takes to get the right approver on the phone.
A simple way to set a decision deadline
The Atlas suggests working backwards. Begin with the target, deduct the slowest recent rehearsal of the technical steps, then deduct a cushion. In its example, 60 minus 33 minus 5 leaves 22 minutes. After 22 minutes of uncertainty, the call must be made, because waiting longer makes the objective impossible even if everything else goes perfectly.
Try it on a small scale. If your booking portal must be back within an hour and the last rehearsal of the technical switch took 33 minutes, your leadership has barely 22 minutes to decide. Write that figure on the plan, and name the person who owns the call plus a stand-in for when they are off shift.
Choosing how much standby to pay for
Not every service deserves the same protection. The Atlas lists four broad strategies, borrowing the names popularised by a major cloud provider:
- Backup and restore: cheapest, slowest. Copies sit idle until needed.
- Pilot light: a minimal core runs all the time, and everything else is started when disaster hits.
- Warm standby: a smaller yet functioning duplicate waiting for users.
- Multi-site active/active: two or more locations answer users at once.
A hot standby, mentioned as a full-capacity cousin of the warm option, sits between the last two. As you move down the list, the cost rises and the recovery time shrinks. A sensible tiering approach, shown in the Atlas, gives the most critical services a one-hour target and a tight recovery point, while less urgent systems get eight hours or even three days. Staff rosters, for example, can usually wait longer than the page where families check visiting arrangements.
If you already read about how telehealth supports follow-up after heart surgery, you will see why the booking and video-visit services belong in the top tier of any such list: families lean on them when it matters. Likewise, a piece on planning specialist care through teleconsultation shows how many steps of a family’s journey now run through online tools that each deserve their own recovery plan.
Switching over and switching back
When the plan is invoked, the Atlas describes a sensible order: first fence off the old system so it cannot keep accepting work, then promote the data to the standby, and only then move traffic across. Skip that first step and two systems may each think they are running the show, logging bookings the other never sees.
One more distinction: a planned switchover, done deliberately while everything is in sync, loses nothing, whereas an emergency failover may drop recent entries that had not been copied across. And returning to the original site afterwards is effectively a second migration, which is why the Atlas urges teams to plan the return trip early rather than as an afterthought.
Common misunderstandings, briefly
- Running across several zones in one region is helpful, yet it is not the same as having a recovery plan for a whole-region problem.
- A provider running your servers does not mean the provider runs your recovery. The Atlas quotes Microsoft’s documentation making the point that recovery does not come built in with its cloud.
- Copying data constantly can copy mistakes just as quickly, so keep point-in-time copies as well.
- Knowing why something broke does not put the service back.
Where a small team can start
You do not need an enterprise budget to begin. List your services, sort them into three tiers, and agree an RTO and RPO for each. Time one real rehearsal, including the decision minutes. Then fix the slowest step and repeat next quarter. The Atlas calls the first maturity level simply “Documented” and the next “Defined”, which is a reminder that every organisation begins on the lower rungs.
For a patient-facing site, the practical payoff is calm. When the portal flickers, staff know who decides, by when, and what the fallback looks like. That clarity is the whole point of DisasterRecoveryOps, and it costs far less than an unplanned scramble. This article is general information about IT operations only and does not address any clinical matter.