Spectra Logic: Taking Tape To Infinity

Many would agree with the following assertions: the IT hardware business is not a good one to be in especially during trying economic times; tape is dead; and only large vendors can be successful in the enterprise-class space, especially at the very high-end. Spectra Logic is a counter example to the first two assertions and, as we will see, is very likely to be a counter example to the third.

Although this story is about Spectra Logic's entry into the high-end enterprise-class tape library market, the broader story is really about how understanding the market dynamics well and then bringing to bear the necessary management and engineering disciplines can result in product innovations that meet even the most stringent customer needs. Capitalism rewards those who understand this dynamic and penalizes those who do not.

And Spectra Logic has been successful in delivering tape solutions, primarily tape library products, which have met enterprise needs for 30 years. They are a hardware company that is also self-funded and manages to grow while remaining debt-free by design. That means the company can do what it feels is in the best interest of the business and its customers without an overemphasis on quarterly results. 

The company has been consistently profitable, establishing a track record that should be reassuring to customers. Although the standard caveat that past results are no predictor of future results applies here, customers want to work with a vendor who is economically sound. All in all, Spectra Logic defies conventional wisdom that hardware is a bad business and tape is dead.

Note that Spectra Logic's history of success is remarkable --although by no means unique -- for a high technology company. Thirty years is a long time in the history of IT. Market trends have changed many times and business models must have been modified many times to adjust to those trend changes (see Digital Equipment Corporation as just one example of what happens when a successful company is unable to change). Spectra Logic has consistently been able to evolve successfully. For example, while it once used Sony tape technology exclusively as a marketing differentiator, it transitioned to supporting LTO tape technology when customers and the market demanded.But that was then, and this is now. The two biggest names that Spectra Logic sees in the tape library business today are IBM and Quantum. A smaller company, Overland Storage, is twisting in the economic winds. StorageTek could be considered as another big name, but the company has fallen upon hard times since it was acquired by Sun (which is, itself, about to be acquired). Yet Sun's StorageTek is the only other vendor that really plays at the high end enterprise space.

With the advent of its T-Finity solution, Spectra Logic has launched itself into the high-end enterprise tape library space, well beyond 10 petabyte (PB) infrastructures. Note, however, that Spectra Logic has already established itself across a broad swatch across mid-market and enterprise-class tape library space, and that T-Finity simply builds on existing proven architectural components for the most part, extending them without bringing any additional risk to the table.  

The most eye-popping number Spectra Logic notes about T-Finity is that the technology can scale to 180 PB (Capacity and performance figures are based on LTO-5). Quite frankly, no one today needs 180 PB of tape media, but what that number says is that while the library is technically finite, from a customer perspective, it may as well be infinite. Customers can continue to add modularly to the same T-Finity library complex for years with all the data remaining accessible within the same system.  

Let's look at more realistic numbers. Spectra Logic feels that there are a number of customers in the 30 to 50 PB range. That leaves only one competitor: Sun StorageTek's SL8500, which maxes out at 106 PB. By comparison, IBM's TS3500 tape library currently maxes out at 10.3 PB. 

That may be all well and good, but who would possibly want that much tape? Although tape usage is generally considered to be horizontal across all industries, some vertical markets command attention, notably high performance computing, media and entertainment, government and healthcare. But what do these companies need all that data for? The answer is not backup. Yes, backup will still be an important application, but that is not what is driving increasing consideration and acceptance of tape storage. The answer is active archiving.Most data that organizations hold is fixed content data that is unlikely to change. For cost and manageability reasons, that information should be moved to an archive. If the data is likely to serve no further operational, management or business use, but still needs to be retained for some business reason, it can be put on tape in what is called a deep archive which may very well be offline.

However, if they choose, organizations can move fixed content to an active archive which is "online" in the sense that a business user can access and manipulate data when and as necessary. Data that needs to be accessed in seconds is best kept on disk. However, much data that can be stored in an active archive - old e-mails, CAD/CAM files, medical images and movies, for example - can be stored on tape.

What about performance? That can be measured for both random access of individual records and for sequential processing. How can random access be performed on tape which is sequential? Well, robotics are used to locate a given tape and move it to a drive, which then has to find the data. Although times can vary the overall time to the first byte of data is about 1 minute 15 seconds. Not fast, but a whole lot less expensive than a disk solution. By the way, in T-Finity and similar solutions, disk storage is used as a cache so the data is fully usable at disk-quality performance.

How does performance stack up against sequential analysis, such as data mining? Surprise, tape is faster. An individual tape drive can stream data at about 140 MB per second compared to 100 MB per second for a disk drive. The bottom line is that huge data repositories will move into active archives where tape will play an important role.

Now obviously Spectra Logic seeks to replace StorageTek at the high-end, and this ambition is attainable. StorageTek is unfortunately only a shadow of its former proud self and constitutes a straw man waiting for Spectra Logic or others to strike a match. The only useful purpose for describing Sun's mismanagement of StorageTek would be to contrast it with Spectra Logic's actions over the last five years, but that sad task can be left to a Harvard Business case study. Spectra Logic now wins against StorageTek on key metrics from footprint to green computing to reliability to TCO across the board. But Spectra Logic not only wants to replace StorageTek in existing markets, it wants T-Finity to dominate the emerging market for high end tape, as well.In either case, in order to be widely recognized as an enterprise-class solution, Spectra Logic has to meet enterprise-class requirements, which the company's T-Finity solution already does:

    * Reliability -- In multi PB solutions, reliability is a must. Spectra offers built in 99.99% reliability, which is typically unheard of for a tape library, but necessary for large-scale tape libraries. It accomplishes this through no single point of failure design points, such as dual robotics controllers.
    * Integrity -- Closely related to reliability of components and part of what delivers four nines availability is what Spectra Logic calls lifecycle management for critical components including the tape library itself (including robotics), tape drives and tape media; Spectra Logic offers drive lifecycle management (DLM), library lifecycle management (LLM) and media lifecycle management (MLM). The purpose of these capabilities is to reduce risk with features such as pro-active notification of upcoming service events and alerting when to replace a tape before it poses a risk to data.
    * Environmental -- From footprint to power consumption, Spectra Logic's modern design has an advantage as compared to older competitive tape library architectures that were designed when floor space and green computing were not the critical issues they are today.
    * Manageability -- Spectra Logic offers a unified management suite that simplifies the administration of large tape libraries (essential to make sure that administration is neither labor intensive nor requires hard-to-find and expensive special skills that would drive up the library's TCO). The suite includes not only a single, well-proven management interface, but also encryption key management, library virtualization, and power monitoring capabilities.

In T-Finity, Spectra Logic has put together a high-end enterprise-class tape library that pays attention to the details that are necessary for delivering true enterprise-class experience. That required not only solid engineering and manufacturing capabilities, but also a clear understanding of what IT buyers expect --  as well as what they truly need -- in an enterprise-class solution. 

But it does no good to deliver a quality product if there is no market for it. And Spectra Logic has clearly identified its target markets in large IT and Federal backup environments. That segment of active archiving where bulk storage of data is necessary and the business user has to be able to access the data "online," and where instantaneous response time is an unnecessary and expensive luxury.

Overall, Spectra Logic seems positioned to prove that you don't have to be a huge vendor to successfully attack a high-end enterprise-class market and thus acts as a counterpoint to conventional wisdom. What the company has accomplished in T-Finity is nothing that their major competitor in this space could not have done, but that is where doing things right rewards the innovator and penalizes the company that refuses or is unable to change.