Tape and object storage are often set against each other, as if a backup strategy had to choose one. In many estates they serve different tiers of the same strategy. Object storage holds the restore points that need to come back quickly and stay protected from tampering. Tape holds copies that are kept for years and almost never read. Understanding what each tier is good at makes it easier to decide how much of the retention schedule belongs on which.
This post describes how the two media fit together in a backup design, and where the boundary between them usually falls.
Different strengths
Object storage is online. Any restore point can be read immediately, many can be read in parallel, and the storage layer can enforce immutability per object. It grows by adding capacity to a running system. Its cost is in the hardware, power and space needed to keep everything spinning or powered.
Tape is offline by nature. A cartridge on a shelf uses no power and cannot be reached over a network. The cost per terabyte of the media is low, and cartridges have a long shelf life when stored correctly. Its cost is in access time, handling and the drives and library needed to read it.
How the two compare as backup tiers
| Characteristic | Object storage | Tape |
|---|---|---|
| Time to first byte on restore | Immediate | Minutes in a library, hours or days from a vault |
| Parallel restores | Many streams at once | Limited by the number of drives |
| Random access | Yes | Sequential, with seek time along the tape |
| Protection from online attack | Immutability enforced by storage | Physical separation when ejected |
| Cost for data read often | Favourable | Unfavourable |
| Cost for data kept many years untouched | Higher | Lower |
| Operational effort | Automated | Media handling, drive maintenance, generation migration |
A common division of labour
A layout seen in many mid-sized and large estates has three parts. Recent restore points, from the last days to a few months, sit on object storage close to production. These are the points used for almost every real restore, and they are held under immutability. A second copy of the same period sits on object storage at another site. Long-term points, the monthly and yearly sets kept for compliance, are written to tape and stored offline.
The boundary between object storage and tape is set by how often data is expected to be read. Anything likely to be needed for an operational restore belongs online. Data kept only because a rule says it must exist for seven or ten years is a candidate for tape.
Restore expectations per tier
Each tier should come with a stated restore time, so that nobody is surprised when an old restore point takes two days to return. Restores from the online tier start at once and are limited by throughput. Restores from tape start with locating and loading the cartridges, then reading sequentially, then writing to production. For a large system that can mean days. This is acceptable for an audit request and not for an outage. Finding the slowest step in a restore describes how to measure where the time goes.
Where tape is under pressure
Several developments have moved the boundary toward object storage. Ransomware recovery needs weeks of restore points available immediately, which pushes more retention online. Backup software now writes incremental-forever chains that depend on random access, which tape does not offer. And tape generations change every few years, with drives typically reading back only one or two generations, so long-retention sets need periodic migration to stay readable.
As a result, some organizations keep only the oldest yearly sets on tape and hold everything else on object storage. Others retire tape entirely and use a second object store for long-term copies. Offsite backup without tape rotation covers that route.
Where tape still earns its place
Tape remains a reasonable choice in three cases. Very long retention of large data sets that are almost never read, where media cost dominates. A requirement for a physically offline copy, either from a regulator or from an organization's own risk assessment. And estates with an existing library, trained staff and working procedures, where the investment is already made. Air gap versus immutability looks at how an offline copy and an immutable online copy differ in what they protect against.
Costing the tiers honestly
Comparisons of tape and disk often consider media price alone. A fair comparison includes the library and drives, their maintenance contracts, the vault and courier, staff time for handling, the cost of migrating between generations and the cost of slower restores. On the object storage side it includes servers, power, space, software and growth. The answer depends heavily on how much data there is and how long it is kept. The cost of another year of retention shows one way to put a number on the long-term portion.
Where ARTESCA fits
ARTESCA is object storage software used as a backup target, deployed on servers the customer runs. In a tiered design it is the online tier: the repository for recent restore points and for the second-site copy, with immutability provided by S3 Object Lock on each object. In compliance mode no account, including ARTESCA administrators, can remove a lock before it expires.
ARTESCA grows within the same deployment from roughly 50 TB to 8.5 PB, which allows the online tier to take on more of the retention schedule over time. It is used as a target for Veeam and other backup software, which can also write long-term copies to tape where that tier is kept.
Drawing the line
A simple way to place the boundary is to list the restores of the past two years with the age of the restore point each one used. In most organizations nearly all of them fall within the last 30 to 90 days. That period, with a margin, is what needs to be online and immutable. What lies beyond it can be kept on whichever medium costs least for data that is rarely read.