The Four Options at a Glance #

NAS/SAN (online disk):

  • Network reachability: always online
  • risk: very high
  • Restore speed: fast
  • Operations: fully automated
  • Role: short-term operational restores only

Tape (LTO):

  • Network reachability: offline once the cartridge is removed
  • risk: low while offline, but exposed in the library and during handling
  • Restore speed: slow (media must be located, mounted, read sequentially)
  • Operations: manual media handling, transport, and storage; no online verification of offline cartridges
  • Role: the legacy offline medium that air-gapped disk systems replace

Cloud (S3/​Azure/​Google):

  • Network reachability: always online
  • risk: high without object lock; moderate with correctly configured compliance-mode object lock
  • Restore speed: depends on bandwidth and egress; large restores take long and cost money
  • Operations: automated, but credential and configuration management is critical
  • Role: geographic redundancy as a supplement, never the primary line of defence

Hardware air gap on disk (Silent Brick System):

  • Network reachability: physically or galvanically separated outside controlled backup windows
  • risk: very low
  • Restore speed: fast (disk-based, random access)
  • Operations: automated separation, online integrity verification, no media handling
  • Role: the modern offline layer for ransomware recovery

Detailed Comparison by Medium #

1. NAS / SAN (Network-Attached Storage) #

How it works:

  • Backup software writes to a network share (SMBNFS)
  • The NAS/SAN is always reachable (24÷7 online)
  • Often with snapshots for fast point-in-time recovery

scenario: obtains admin rights, finds the NAS credentials in the backup software configuration, connects to the share, and deletes or encrypts the backup data.

resilience: very weak.

  • Network-reachable, with credentials stored locally
  • No enforced immutability (an admin can delete everything)
  • Snapshots live on the same system as the data

When to use: Only for short-term operational restores (24 to 48 hours). Never as ransomware protection. Always combine with an offline layer.

2. Tape (LTO) #

How it works:

  • Data is written sequentially to LTO cartridges
  • Cartridges are removed from the library and stored offline
  • Historically the default for offsite and long-term copies

scenario: A cartridge sitting in a vault is unreachable for ransomware. But cartridges still in the library, and the backup server controlling the library, are not protected. Attackers who compromise the backup server can erase or overwrite loaded media and sabotage future jobs.

resilience: offline cartridges are protected; the overall process is fragile.

The operational price of tape:

  • Slow restores: media must be retrieved, mounted, and read sequentially; large restores take many hours to days
  • Manual media handling: removal, labelling, transport, and storage are human processes, and human processes fail silently
  • No online verification: an offline cartridge cannot be checked; you discover unreadable media during the restore, which is the worst possible moment
  • Generation and drive compatibility management over years

Assessment: Tape delivers the offline property, but at the cost of slow recovery and error-prone manual processes. This is precisely the gap that on disk closes: the same offline protection, with fast random-access restores, automated separation, and continuous integrity verification. For organisations still running tape, replacing it is one of the most effective resilience upgrades available.

3. Cloud (AWS S3, Azure Blob, Google Cloud) #

How it works:

  • Data is transferred via HTTPS to a cloud provider
  • Object lock () can be enabled
  • Scales to very large volumes

scenario: finds cloud credentials locally (in a config file), connects to the bucket with stolen credentials, and deletes or re-encrypts the backups. Governance-mode object lock can be bypassed with the right stolen permissions.

resilience: moderate to weak.

  • Credentials are reachable on local systems
  • Object lock in governance mode is bypassable; only compliance mode enforces retention strictly
  • Large restores are slowed by bandwidth and burdened by egress costs

When to use: Geographic redundancy and disaster recovery as a supplementary layer, for non-critical data, or as an additional copy. Never alone for ransomware protection, and never as the primary strategy: the last line of defence belongs on premises, under your physical control.

4. Hardware Air Gap on Disk (Silent Brick System) #

How it works:

  • Backup data is written over a secured connection during controlled windows
  • After the backup, the storage is separated from the network: with Silent Brick Pro, the bricks are physically removed from the Controller X (a true physical air gap); with Silent Brick Max Air, galvanic separation disconnects the storage electrically, with no physical removal needed
  • Data integrity is verified by the system, so you know the copy is restorable before you need it

scenario: has admin rights and attempts to reach the backup target. Outside the backup window there is no data path. Access is impossible, and the backup stays intact.

resilience: very strong.

  • Physically or galvanically isolated, not addressable from the network
  • No credentials that can be stolen to reach offline copies
  • Automated separation removes the human error of manual media handling
  • Disk-based: restores start immediately, with random access instead of sequential tape reads

When to use: As the offline layer for critical systems (domain, ERP, email, file services), daily backups with fast RTO requirements, and as the recovery foundation required by frameworks such as NIS2 (Directive (EU) 20222555 lists backup management and crisis management among the mandatory risk measures) and for the financial sector.


No single medium is optimal. Recommended strategy by layer:

Tier 1: Hot Recovery (Fast Operational Restores) #

  • Medium: NAS/​snapshots
  • RTO: minutes to a few hours
  • Retention: 48 hours to 7 days
  • Purpose: fast local recovery from everyday failures
  • protection: no

Tier 2: Offline Layer (Ransomware Protection) #

  • Medium: on disk
  • RTO: fast, disk-based restores
  • Retention: weeks to unlimited
  • Purpose: guaranteed recoverability after a network-wide compromise
  • protection: yes

Tier 3: Geographic Redundancy (Disaster Recovery) #

  • Medium: cloud or a second site (supplementary copy)
  • RTO: hours, depending on bandwidth
  • Retention: 90 days to 1 year
  • Purpose: protection against site-level disasters
  • protection: partial; depends on object lock configuration

Example implementation for critical systems:

  • Continuous: snapshots to NAS (hot recovery)
  • Daily: backup to the (Tier 2)
  • Weekly: copy to a second site or cloud for (Tier 3)
  • Monthly: full recovery test from the air gap layer

For regulated long-term archiving with legally mandated retention, add hardware storage (Silent Cubes) as a separate archive layer; that is an archiving task, distinct from backup.


Decision Guide: Which Medium for Which Data? #

  • Is the data critical for business operations? Then it needs Tier 1 (fast restore) plus Tier 2 (air gap). Add Tier 3 for site-level risk.
  • Is the data non-critical? NAS plus a supplementary offsite copy is sufficient.
  • Do you need fast recovery after a ransomware attack (RTO under a day)? Then the offline layer must be disk-based. Sequential offline media cannot deliver that.
  • Do you need compliant long-term archiving (6 to 30 years)? That is a separate requirement: hardware archive storage (Silent Cubes), not the backup layer.
  • Are you still rotating tapes? Evaluate replacing the tape process with an automated : same offline property, faster restores, no media handling, verified integrity.

Frequently Asked Questions #

Should I use only one medium? No. Redundancy is the point of the 32110 rule: at least two different media, one copy offsite, one copy offline or immutable, zero errors in the restore test.

Is tape still a valid choice? Tape provides the offline property, and that is its one remaining argument. Everything else speaks against it: slow sequential restores, manual handling, no online verification of offline media. A  on disk provides the same protection without these drawbacks. That is why organisations replace tape with air-gapped disk systems.

Why is cloud alone not enough? Because credentials can be stolen and large restores are slow and expensive. Cloud works as Tier 3 (geographic redundancy), not as primary ransomware protection. The primary offline layer belongs on premises.

What does a secure multi-tier solution cost? That depends on data volume, retention, and RTO targets. The decisive comparison is not media cost per terabyte but total risk: industry reports consistently put the full cost of a ransomware incident (downtime, recovery, forensics) at a multiple of any backup investment. Request a sizing consultation for concrete numbers.


Further Resources #

→ Logical vs. Physical : The Difference (/en/blog/logischer-vs-physischer-air-gap/) → Hardware : Comparison for IT Decision-Makers (/en/blog/hardware-air-gap-vergleich/) → Why Cloud Backups Provide No Real Protection (/en/blog/cloud-backup-ransomware-schutz/) → The 32110 Backup Strategy (/​en/​blog/​3 – 21 – 10-backup-strategie/) → Silent Brick System: Hardware for Protection (/en/produkte/silent-brick-system/) → Request a Demo (/​en/​kontakt/​demo/​)

Disclaimer

This article was written by our editorial team and edited using AI. It provides a general overview and does not constitute legal advice – we recommend seeking professional advice for your specific situation.