A few days ago, the Linear Tape Open (LTO) consortium -- whose member include Hewlett-Packard, IBM, and Quantum -- officially released specifications for the next generation of the LTO Ultrium standard, LTO Generation 5. This may sound a little retrograde. With all of the incredible advances made in other storage technologies, can tape maintain its relevancy into the future? Can Virtual Tape Libraries (VTL), removable disk, or the cloud replace the need for constant management of tape-based backup and archiving systems?
To answer these and other questions, let's look at what tape does well and why it has been an indispensable part of the data center for the past 30 or 40 years. The pinnacle of existing tape backup technology currently provides three main benefits.
[ Get the big picture on backups in W. Curtis Preston's Backup Deep Dive Report. ]
First, tape is able to store a lot of data in a very small package. Second, currently available tape drives can deliver sequential transfer speeds that rival the fastest hard disks available. Finally, tape media can be removed from its drive, shipped, stored, and used years later.
As Andrew Tanenbaum, the famous computer science professor and textbook author, once said: "Never underestimate the bandwidth of a station wagon full of tapes hurtling down the highway."
Over the years, tape backup media has maintained parity with the disks it backs up. The current generation of the LTO standard, LTO-4, can natively store 800GB (uncompressed) on a single tape that occupies a little over 14 cubic inches. The LTO-5 standard will roughly double that native capacity to around 1,500GB per tape, yielding a total density by volume of about 107GB per cubic inch. The largest standard internal 3.5-inch SATA disk can hold about 2TB in a package occupying about 23 cubic inches, for a density of about 86GB per cubic inch. If you are planning on packing the entire Internet into the back of an Oldsmobile, tape is still your best option.
Disk may not be able to match tape in density, but surely it must surpass tape in performance, right? Think again. The fastest current-generation LTO drives are capable of around 120MBps total throughput, while the LTO-5 standard extends this to 140MBps (uncompressed). Your average high-capacity SATA disk is lucky to get 35MBps. You'd need to stripe together four or five disks to get the same throughput.
The only other storage technology that can match this kind of sequential data throughput on a single media is solid-state disk (SSD), and at current prices and capacities, replacing all of your tape capacity with SSD would be, well, dumb. Tape wins again.
But the fact that tape can be removed, shipped, and stored for long periods of time is truly what makes it so popular. It's difficult to trust that a device with as many moving parts as a hard drive can endure being shipped to an offsite storage vault -- or sit in a cabinet for 10 years until you need it to run again. Tape, if handled correctly, can last 15 or more years and provide a bulletproof archive capability.
Given all of the reasons why tape dominates backup technology, why would anyone want to rid of it? Because it's inconvenient. To use your backup as a hedge against fire or natural disaster, backups need to be made onto tape at least once a day and then either stored offsite or in a fireproof safe. That means that some poor sap has to remember to go over and pop the tapes out and carry, ship, or drive them somewhere. Every day. Always. In an age where VPNs and remote management tools make it almost unnecessary to get out of bed to run an enterprise network, tape backups are one of the few remaining holdovers that require someone to physically go do something. How distasteful!
Another reason most administrators tend to dislike tape is the amount of time that it can take to perform restorations of small amounts of data. While backups are generally very fast -- the disk being backed up is usually the bottleneck, not the tape drive -- restores of a few tiny files can take a few minutes. This is due to the linear nature of the media: The tape drive has to spin its wheels to find the tiny little bits that you asked for. Hard disks are random-access devices that seek data in milliseconds, not minutes.
So what are the real tape alternatives? Virtual Tape Libraries are great pieces of hardware, but they aren't meant to replace actual tape libraries. VTLs just supplement tape as an immediately available restoration source and backup cache for tapes that are used for archival purposes. Even if you take advantage of some of the great replication features found in some VTLs, you still can't call that an archive, even if it is offsite. Just the fact that the remote VTL is connected to a network means that anyone with the right access, such as a disaffected administrator, can remotely destroy all of the data with very little effort.
In my opinion, the cloud presents the most interesting option. The idea of shunting all your backup data to a third party isn't new; organizations have been vaulting their backup tapes with third parties for decades. Doing it constantly and automatically over the Internet is an enticing concept.
Yet it doesn't take a rocket scientist to realize that the Internet will have to grow radically if everyone decides to replace tape with cloud services. To transfer a single LTO-4 cartridge's worth of data across a network circuit within a 24-hour window would require more than 75Mbps of dedicated bandwidth -- more than most of us have at our disposal. Can the cloud accommodate backups of our most mission-critical data? Sure -- but that's about it. Plus, if it's worth backing up, it's probably also worth archiving securely.
How can you replace tape? Really, you can't -- yet. However long tape's history may be, it's not over. If tape capacities keep pace with hard disk capacities and the growth of our data sets, tape will be around for a long, long time.