The number of cartridges in a cupboard is not the number of media a backup rotation needs. A workable rotation must leave the team able to run the next backup while a recent, verified copy is held elsewhere, an earlier set is being tested, or a handover has been delayed. This is capacity planning, but it is also custody planning.
Begin with the recovery sets, not the catalogue
First define what a backup set means for the protected workload. It may include files, a server image, a virtual machine, application data, configuration exports and recovery dependencies. The size of a production disk is not automatically the size of the backup set: some volumes may not be protected, while an essential configuration may sit outside the obvious data location.
Measure a representative result from the actual backup process and record the assumption used for growth. Full, incremental and synthetic operations can occupy storage differently, and compression or deduplication cannot be treated as a guaranteed ratio. A prudent planning figure, reviewed over time, is more useful than a precise-looking guess.
Before counting media, confirm whether one medium holds the required set with headroom. If one set needs several media, that multiplier belongs in every part of the rotation plan.
Give each medium a job
A good plan makes the status of every medium obvious. At any point, media may be:
- scheduled for the next backup;
- holding a local recovery set that is still retained;
- under separate off-site or offline custody;
- reserved for an exception, replacement or recovery test;
- awaiting a documented decision before reuse.
This avoids the most common planning error: counting only the media that are currently attached. The medium held off site is doing an important job precisely because it is not available to the day-to-day cycle.
Convert policy into a count
Write down the policy in operational terms. How often is a new set created? How many sets must remain recoverable? Which set leaves the site, for how long, and what happens to weekly or monthly copies? How many spare media are required when someone is absent or a set needs investigation?
One planning expression is:
media required = media per set × sets retained at the same time + separately held media + operational reserve
It is deliberately a planning tool, not an instruction for a backup application. The product and software may manage chains, retention and rotated repositories in specific ways. Those settings must be validated in the current vendor documentation and against the actual environment.
Work through a realistic example
Consider an organisation that creates one daily set, keeps five weekday sets, sends the latest verified set to another location for a week and keeps one spare medium. Five media alone are unlikely to keep this arrangement running: the separately held copy cannot safely be reused simply because the next day has arrived, and the spare protects the schedule if a handover fails.
The answer changes if the set occupies more than one medium, if a weekly set is retained separately, or if a monthly copy has a different retention rule. For that reason, turn the calculation into a calendar before purchasing anything. Give each medium a permanent identifier and record its contents, location, custodian, verification status and earliest permitted reuse date.
Test the calculation against normal disruption
Ask what happens when the planned medium is missing, a job has failed, a person has not returned an off-site copy or an older set is required for investigation. A robust plan protects the last known-good copy while the exception is resolved. It does not overwrite it because the next job is due.
This is also where an operational reserve earns its place. A spare medium is not waste when it prevents the team from weakening retention under pressure. The same applies to a medium temporarily reserved for a restore test: the test must not silently remove the organisation’s only separated recovery option.
Make restoration the final check
Choose an older, representative set rather than always restoring the latest one. Use the register to locate the medium, obtain it through the normal custody process and recover the agreed scope. Measure the full time: request, location, transport, access, recovery and business acceptance. Record every dependency that was needed.
If the result misses the recovery objective, the remedy may be a revised rotation, clearer custody, more capacity or a different recovery design. The crucial point is to correct the evidence, not only the spreadsheet.
Keep the decision connected to the wider strategy
Media count supports recovery point and recovery time objectives; it does not define them. More media cannot compensate for an inadequate backup frequency or an untested restore. Equally, a carefully retained set has little value if nobody can find and use it.
Use the offline media rotation guide, 3-2-1-1-0 rule and restore-testing checklist to complete the process. Refer a reader who needs product selection, capacity validation or architecture review to the Tandberg Data main site.
Share this guide
