Tape isn’t as “sexy” (from a tech perspective) as solid-state drives (SSDs) or the cloud, and the number of articles written about tape is far lower than for those technologies. Even dedupe probably gets more ink, although it’s been
As Todd Erickson’s article “National Geographic chooses tape archiving to store PBs of content” on SearchDataBackup.com points out, tape is still a very viable solution for data archiving, especially in the largest environments such as National Geographic’s. I spoke with Kyle Knack, National Geographic’s director of infrastructure systems, at a Spectra Logic event last fall and wrote a short piece on the company's archiving needs. At that point it was pretty clear National Geographic had no real alternative to tape; there was just too much data. A petabyte of new files each year obviously puts the company out of the market for traditional disk storage. While there’s still no alternative for the largest archives, I think the take-away from this article is that even smaller environments would be advised to consider tape archiving. And, VARs that have ignored tape would be advised to get comfortable with this technology (again) and bring it up more often.
Reference data, unlike backup data, doesn’t need to be modified regularly, especially the raw copy (or copies) that users put into their libraries first -- even those much smaller than National Geographic’s. When you consider that most of these data sets consist of image-based files (pictures, sensory data and videos), they typically don’t dedupe very well and usually get large or very large in size. The retrieval patterns for these kinds of files are less random and more streaming in nature. All of these factors support a linear recording format such as tape.
The knock against tape in non-backup applications has always been latency. But in the final analysis, the viability of a technology is determined by how well it functions and how much it costs. For most archives, especially those with larger files, tape systems can have plenty of performance. When paired with spinning disk (or flash) to serve up the first few minutes of a video or lower-resolution images for browsing, a tape archive can support many use cases. When cost is factored in (including floor space, power and cooling), plus expandability, tape is an even better choice.
Even questions about tape’s long-term viability are being addressed by library- and drive-level routines that track media errors, head wear and drive duty cycle. This kind of intelligence embedded into the tape infrastructure enables the prevention of recovery errors that plagued many early tape backup users.
File formats abound, and file sizes keep growing. As Erickson reports, National Geographic has to accommodate 25 file-based media formats and adds more than a petabyte each year. Even in smaller environments the problem is the same: how to keep saving one or two (or more) copies of everything. In this situation, tape is unsurpassed. LTO-5 cartridges currently hold 1.5 TB (native), and the LTO roadmap extends to LTO-8, which is slated to hold more than 12 TB per LTO cartridge, uncompressed. This puts the data density of tape (terabytes per data center floor tile) at multiples of even the largest disk drive arrays.
Don’t be afraid to suggest tape
National Geographic said tape costs it one-tenth of what even its lowest-price disk drive capacity does, and that doesn’t factor in power, cooling and floor space. Even for smaller shops, this kind of savings is compelling. Tape’s essentially unlimited expandability makes it a good solution for IT managers charged with storing a copy of everything. When you consider recent developments in the technology surrounding tape, such as the Linear Tape File System (LTFS), tape media lifecycle management and its use in “big data” environments, pitching tape archiving can be a way to get in the door at a new account. Who knows, maybe tape will become sexy again.
Eric Slack is a senior analyst with Storage Switzerland.
This was first published in January 2012