Tape vs. disk storage: Why isn’t tape dead yet?

It might surprise some that despite the popularity of disk storage, the amount of data currently being placed on tape continues to rise.

For example, 2017 saw the shipment of more than 1 million petabytes in LTO tape capacity – roughly five times what was shipped in 2008. At the same time, the tapeless backup and recovery system has become the norm. Every backup expert that I know recommends disk as the primary target for backup and recovery.

How then to explain the increase in the amount of tape being shipped? The answer might be surprising, but first let’s take a look at what tape is good at – and not so good at.

Tape is cheaper than disk

Even with all the advancements in deduplication – which is primarily used on disk – tape is still cheaper per gigabyte than tape for a few reasons. Tape allows you to separate the medium from the recording device, which allows you to buy a handful of tape drives and thousands of tapes. Those thousands of tapes also do not need power and cooling to maintain their data. In fact, some have suggested that even if disk was free it would still cost more than tape due to the power and cooling savings alone. The fact that tape is less expensive than the alternatives is the main reason we’re still talking about it.

Tape is better at writing data

When writing bits to storage, there is something called an unrecoverable bit error, which is when a device stores a one instead of zero or vice versa and it isn’t fixed with error correction. What many people do not understand about tape is that it actually has a better bit error rate than any other recording medium.

LTO-7/8 and Oracle T10000 tape drives both have unrecoverable bit error rates of 1:10¹⁹, which is roughly one error every 1.25 EB. Enterprise-class disk drives have an error rate of 1:10¹⁵, or an error every 125 TB. That essentially means that tape is 1000 times better at writing ones and zeros than the best disk drive. (In case you’re curious, SSDs range from 1:10¹⁶ to 1:10¹⁸, so tape is also 10 times better than SSDs at writing data.) It is possible to address the differences between disk and tape using higher-level checks of the data, but the fact remains that tape is simply better at writing data than disk.

Tape is better at holding onto data

The most misunderstood thing about tape is how good it is at holding onto data, which is actually a matter of basic recording physics. All magnetic bits are subject to degradation over time; the only question is how quickly it will happen. Those familiar with the technology refer to something called “blocking energy” that prevents a magnetic bit from changing its position over time, which is expressed in the formula KuV/kt.   The two relevant values in this discussion are the volume of the magnetic particle (V) and the temperature at which it is stored (k). The larger the magnetic particle and the lower the temperature at which it is stored, the better.

Compared to disk, tape drives have very large magnetic particles and tape is stored at ambient temperature.  Disk drives have much smaller magnetic particles and they are constantly running at much warmer temperatures than a tape is typically stored at. The result is tapes have much higher blocking energy than disk drives, and can hold onto data for much longer periods of time. Specifically, the data says not to store an individual piece of data on a disk drive for longer than five years, but you can store the same file on tape for as long as 30 years.

Disk has a much better random-access time

People have grown used to very quick restores of their data via things like Time Machine and continuous data protection. Those things are all made possible by the much lower random-access time of disk. Disk is literally a thousand times faster than tape at getting to the first block of data. Depending on how you are using tape, this could add a minute to restore, or it could add hours. If your backup software does many seeks during a restore, tape’s access time can kill you.

Disk is better at matching speed than tape

The biggest problem with tape is actually that it cannot keep up with the speed of most backups; it is actually too fast. The compressed transfer rate of standard LTO-8 tape-storage technology is 750 MB per second. It’s really good at going that speed, and it’s really lousy at going less than that speed. The problem is that most backups are incremental backups and they provide a few megabytes per second, which is simply incompatible with tape.

Not only can disk run at a few megabytes per second – or a few kilobytes per second if that’s what you need – it can actually run hundreds of backups simultaneously at that speed while still storing each backup in a separate location. Tape cannot do that. This makes disk much more appropriate for backup than tape.

Why all the tape?

If people aren’t using tape for their backup systems, who’s using all that tape? The answer is secondary storage. Tape is increasingly being used to store large amounts of reference data that needs to be reliably held for long periods of time as cheaply as possible. Companies like entertainment and biotech companies with big files (where the random access time issue really isn’t relevant) are buying huge tape libraries and filling them with thousands of tapes.

The thing that will surprise most people is that cloud vendors are using tape for the same purpose. Why do you think access times for some of the less expensive storage services are so long?  Some of them are absolutely using tape in the back end to provide reliable long-term storage of your infrequently accessed data. What many people think of as the future is built on what many people think of as the past. Think about that the next time you try to say the tape is dead.