Anyone who has priced a private charter knows that speed and certainty cost money, and that you rarely need the maximum of either for every trip. That trade-off is also the core of DisasterRecoveryOps, the habit of readying IT services for a serious failure. When the Ops Atlas lays out the ways an organisation can prepare, it presents four strategies that line up neatly with ways of securing aircraft capacity. This comparison walks through them side by side and weighs what each buys you.
The four DisasterRecoveryOps options at a glance
In the Ops Atlas entry covering DisasterRecoveryOps, the strategies appear with names borrowed from cloud providers: backup and restore, pilot light, warm standby (with a hot-standby variant running at full capacity) and multi-site active-active. Here is the aviation reading of each.
- Backup and restore resembles phoning a broker after your aircraft is grounded and starting the search from scratch.
- Pilot light is like holding a standing arrangement with an operator, with crew and aircraft not yet allocated but the relationship live.
- Warm standby compares to an aircraft positioned and prepared at a nearby airport, smaller than the main plan but ready to scale.
- Active-active is two aircraft flying the schedule at the same time, either of which could carry the whole load.
Option one: backup and restore
This is the cheapest and the slowest. You keep copies of everything and rebuild after the event. Its strength is simplicity, and its weakness is that recovery can take hours or days. In the Atlas’s sample tier matrix, the least critical services accept a recovery target of 72 hours and a 24-hour data-loss window, and this strategy fits them.
The charter parallel is the occasional traveller who books whatever is available on the day. Prices swing, and so does availability, which is why a pricing piece such as our breakdown of what drives the cost of a private jet charter is useful reading for anyone relying on last-minute supply.
Option two: pilot light
A pilot light keeps the most essential core of a service running in a second location, so that it can be scaled up when needed. It costs little to maintain, and recovery is faster than from a cold start, though still far from instant. In the sample matrix, a middle tier with an eight-hour target and a one-hour data-loss limit uses this approach.
The comparison with a retainer arrangement is apt. You pay a small, steady amount for priority, but you still need time to mobilise everything else. The risk is the usual one for a skeleton setup: when it is finally called on, it may have drifted from the live system, with stale images or quotas that were never raised.
Option three: warm and hot standby
Warm standby runs a reduced-capacity copy of the service all the time. Hot standby, the full-capacity variant, can take over almost immediately. This is where the Atlas places its most critical tier: a one-hour target, a five-minute data-loss limit, and a planned switchover rehearsed every quarter.
The cost is continuous, because you are paying for capacity that mostly idles. That raises a question operators of charter fleets know well: how to make use of the aircraft sitting positioned for resilience. Our explanation of how empty-leg flights come about and why they are sold describes one answer on the aviation side, and the technology equivalent is using the standby for rehearsals and read-only work so the spend is not wasted.
Option four: active-active
Running two full sites together gives the fastest recovery and the highest bill. It also brings design complexity, since both sites must agree on shared data. The entry’s remark on replication bears repeating here: copying everything everywhere does not protect you from a fault that is itself copied, such as a bad update or corrupted data, so older point-in-time copies are still needed.
The Atlas offers a cautionary example. On 19 July 2024 a defective content update took down Windows hosts, with Microsoft’s own estimate standing at 8.5 million devices. Standby sites were no help, since every site ran the same flawed update. That is the aviation equivalent of a systemic defect found across a whole fleet, where having more aircraft does not solve the problem.
Reading the costs
The Atlas includes concrete prices for one cloud tool, which makes the point that resilience has a price tag. Its listing for Amazon’s traffic-switching service shows a Region switch plan priced at $70 monthly, plus a routing-control cluster billed per hour at $2.50, which the sheet puts near $21,900 across a year. Whatever the exact figures turn out to be when you check, the structure is the lesson: some protection is a rounding error and some is a line item that must be justified.
The Atlas also warns against demanding zero downtime and zero data loss from everything, which is a stance every charter buyer understands. Nobody wants a standby aircraft for a trip that could move by a day.
What the fine print adds
Two cautions from the entry apply to every option. The first concerns automatic failover triggered by health checks: the Atlas says it should be used with caution, because a brief glitch can trigger a costly switch that nobody wanted. The second concerns the way a switch is carried out. Doing it by hand through a console, the so-called click-ops method, is listed as an anti-pattern, since steps that live in someone’s memory rarely survive a stressful morning.
For operators of aircraft the analogy is direct. A positioning flight launched on a sensor reading alone, without a human reviewing the whole picture, can create more disruption than it prevents, and a procedure that depends on one dispatcher’s recall is a procedure waiting to fail.
How to choose between them
- Start from the consequence, not the technology. Ask what an hour of outage costs for each service, then rank them.
- Match the event to the strategy. Spare capacity within one building or zone may handle the loss of a rack but not the loss of a region. The Atlas notes that multi-zone designs may need only restore-from-backup for single-site loss, while regional events need strategies that span regions.
- Add the decision to the clock. Whichever strategy you choose, the human minutes of noticing and deciding count against the target. The Atlas’s example drill missed a 60-minute target by four minutes, entirely because of those minutes.
- Plan the return. Every strategy ends in a failback, which the Atlas describes as a second migration.
- Rehearse. A well-chosen strategy is worth little if nobody has timed it.
The verdict
No single strategy wins, and the Atlas does not pretend otherwise. The pilot light and warm standby tend to give the soundest balance for services that matter but not catastrophically; backup and restore suits the rest; and the two-site model belongs to the very few functions where minutes of interruption are intolerable. For a charter business, think of it as matching reserve capacity to the route: modest for the routine, generous for the one flight you cannot miss.
A last practical suggestion: write the chosen strategy for each service on a single page, with its target time alongside and whoever is authorised to order the switch. Revisit the page after every drill. If the numbers on it keep being missed, either the target was unrealistic or the strategy was too modest, and both are worth knowing before an incident rather than during one.

