Disk versus tape: A retrospective with an eye on the future

VOXDC_Header_Multi_Colors.png

I used to be a night operator running “data processing” jobs. The jobs were batch and predominately tape-based (this detail tells you something about my demographics—but I digress). The tapes we used were 2400’ reel-to-reel at a density of 1600 bits per inch. The density was so low that you could actually see the bits on the tape![1]

Technology has changed substantially since then, and today there are a few reasons why you might choose on-premises disk storage over tape in the modern data center.

Trash the Tape - 1.pngIBM tape drives—2400' reel-to-reel.Why use tape?

  1. The economics of the media. Tape media has the lowest cost of any magnetic storage media. In the past, a 2400’ foot reel of tape held 45 megabytes, whereas today, an LTO-8 cartridge capacity is about 15 terabytes, uncompressed—a capacity factor of 333,000 times more density at roughly the same price!

    Of course, there’s a lot more investment than just the media—there are tape drives, and if you manage enough tapes, you’ll want an automated tape library system.  While these details add procurement costs, tape maintains fairly low capital expenditure, albeit at the cost of increased operational expenses.
  1. The “air gap.” Workloads supported by tape have changed a lot in the 30 years. Today, most tape systems are used for data protection: backup, archive and long-term retention (LTR). Tapes have a distinct advantage—once written, the tape can be physically removed from the data center and stored securely off-site. The “air gap” is so-called because once removed from the premises, these tapes cannot be touched by the enterprise’s infrastructure, which helps prevents corruption and protects against disasters like ransomware attacks.

  2. It’s hard to change. Many data protection processes started with tape based on the advantages outlined above. While the technology has evolved over time with new, improved drives, formats and management software, it’s hard to change the operational processes; tape-based workflows often remain the same. In fact, the inertia to change can often override the advantages of new ways of doing things; it’s the, “that’s the way we’ve always done it, and we’re not changing now!” syndrome.


Why use disks?

  1. Data availability and lower RTO. By far the major advantage to on-premises disk storage for archive and LTR is the reduced recovery time objective (RTO). The disks are always online, and the data is immediately available—even with cost-optimized disk enclosures that use lower-cost (and thus lower-performing) interfaces. Having data online has also become especially important with regulations like GDPR, where data subjects have the right to see and control their personal data. (Try searching a tape for an individual record, reading and updating it, and rewriting it—just to remove/anonymize a specific chunk of personal data.)

Further, disk-based data is available for full-data sweeps, for example as input to data classification and analytics applications that need access to nearly all of the data in an archive set.

  1. Reduced operational expenses. Unlike tape systems that require human or mechanical intervention to make the data available, there are no substantial additional expenses—other than power and cooling—to keep the data immediately accessible. As a result, disk-based systems generally have a higher capital expenditure but reduced operational expenditures.

  2. Smooth the transition to cloud. Nearly every on-premises disk-based system has some provision for policy-based migration to cloud. If you’re thinking, “Some of this data could be stored on my favorite public cloud,” then a disk-based system is a solid stepping-stone to making that migration.


Comparing disks vs. tape:
  Screen Shot 2019-06-10 at 10.00.38 AM.png

In the end, we come up with two different types of storage media—and here’s the important part—those different types of media should be used for different types of data. Cost-of-ownership takes high priority but has to be balanced with the value of having that data readily available. Data that a few years ago was labeled “cold and destined for off-site cold storage” can today be the lifeblood of analytics applications—applications that may increase an organization’s competitiveness.

Conclusion

If you skipped the storytelling and simply want to get to the bottom line: the answer—it depends on the needs of your organization and its data; how often you access data and its value, as well as your data recovery service level objectives. As discussed on tape—it also depends on your willingness to evolve time-honored data protection processes. And keep in mind the value of archived data is increasing every day—as disk systems’ costs go down and data value goes up, it makes a case for some amount of on-premises disk-based backup/archival infrastructure.

It may even be time to compare cost-of-ownership of tape vs the “always on” advantages of on-site disks. We mentioned tape has lower capital expenditures but higher operational costs; the inverse is true for disk.  In fact, for many workloads the total cost of ownership over time is nearly the same. Notably, Veritas has a quick-and-dirty TCO tool that can get you started with that comparison.

Often, the best approach is a hybrid, driven by identifying which data is destined for the cold, damp cave of long-term retention, and which is of value when it’s always online. And don’t forget analytics applications that periodically require a majority of the data, or data that you might want to tier to cloud.

Fortunately, Veritas can help either way. The comprehensive Veritas data protection portfolio helps you assess and understand the meaningful details of your organization’s data so you can provide automated policies to reduce operational expenses while maximizing protection. Veritas’ long-term retention storage systems, like NetBackup and Access appliances, can be an economical part of a balanced approach. And if tape is part of that protection hierarchy—no problem!

[1] Our hardware engineers could develop the tape, like developing film, by dipping a section of tape into a solution of alcohol and fine iron particles.  The iron particles would stick to the magnetized 1-bits, and with a magnifying glass you could actually see the bits.