A Storage Operating System for the Masses?
Real storage innovation may be happening in software, not hardware
Anyone who has purchased a storage array recently is aware that these are more than simply boxes of disk drives. From the most inauspicious USB-attached or SCSI direct attached storage (DAS) array to the most exclusive network attached storage (NAS), iSCSI or FC target, most products proffer at least some "value-add" features in their "heads."
While there has been some standardization on a standard backplane and bus, especially in lower-cost products (usually a PCI motherboard you might find in servers and PCs), the big brands prefer to create their own backplanes and proprietary data-pathing schemes. This has the two-headed consequences of possibly adding to the performance of the box itself (a good thing) while probably obfuscating universal management schemes (a not-so-good thing). The real innovation, however, is occurring in software, not hardware.
Vendors have taken to building their own storage operating systems onto the boxes, sometimes hardwired in ASICs, but more often than not as software running in a kernel operating system or a full blown server OS. Software approaches are increasingly preferred because of the ready availability of "universal" operating system platforms, whether from big OS vendors such as Microsoft (with its Storage Server) or from the open source community (Linux or BeOS).
Many small vendors are leveraging these low-cost, software-centric approaches to wreak havoc on the large box makers, as evidenced by market-share losses to big-brand vendors from relative newcomers such as Adaptec, Compellent, 3Par, LeftHand Networks, EqualLogic (now Dell), and a few others.
For consumers seeking a plug-and-play solution with special features including the latest functional darlings of the industry -- de-duplication, compression, thin provisioning, data encryption, automated hierarchical storage management (usually from SAS to SATA drives inside the array), specialized read or write caching, special RAID, snapshots, point-in-time mirror splitting, and maybe off-box replication -- the storage OS is the important thing. Today, practically anyone can sell a box of Seagate drives with a RAID controller; it's the integrated feature/function set that discriminates one box from another.
On the other hand, it is this variability in the storage OS that is leveraged by vendors to lock in the consumer and lock out the competition. Functions such as off-box replication only work between two boxes from vendor A. In the worst situations, data itself is held hostage either by proprietary encryption processes or by application programming interfaces that enable data to be written to the vendor's box but not removed from it -- except with the vendor's proprietary data movers.
I pay attention to developments in the storage OS, and so should you. Ultimately, the path to universal storage management, business continuity, and even data management requires that we wrangle some sort of centralized control over the peculiarities in the various storage OSes in our heterogeneous infrastructure.
About a month ago, IBM bought a little company called XIV. To hear Dan Powers, vice president and integration executive with Big Blue, tell it, the XIV acquisition was a result of IBM's ongoing tactic to put merger-and-acquisition teams on the road hunting worldwide for companies whose products fit with IBM’s strategy. The Israeli firm had a "storage architecture" that the M&A folks liked a lot.
"Current storage architectures," Powers said, "are 15 to 20 years old. It is getting harder to add new features and functions while supporting legacy designs." XIV had created a storage architecture based on industry-standard chips and backplanes, plus it was built on Linux OS software (Powers was previously the author of IBM's Linux strategy) and SATA standards, which added to its appeal.
Powers gushed over the new acquisition: "The magic is in the software, so you can add features cost-effectively." He was referring to the XIV software's support for thin provisioning, flash copying, and a parallelized processing approach that "spreads data across all of the disks included in a grid (comprised of multiple arrays) and … that enables the rebuild of a 500 GB drive in 15 minutes or less without any performance impact."
Essentially, he reported, XIV technology will make storage "a non-issue from the standpoint of ease of administration and ease of setup." Its compatibility with x86 and PowerPC chip and motherboard standards likewise meant that IBM could begin leveraging the technology on its own branded arrays immediately. It sounded too good to be true, and Powers promised a follow-up meeting to schedule a deep dive into the architecture. We will report on it here when it happens.
At first, it looked like IBM was taking a different path to success in storage. By offering a universal storage OS that did not require the purchase of proprietary hardware, the company seemed positioned to break with the pattern of its brand name peers. In essence, consumers could buy hardware at street prices and just license the software from IBM to give it a functional personality. That would be a great way to build storage and to buy just those features and functions that the IT manager needed to serve applications.
Alas, that is not the case. "IBM has no plans currently to OEM the software," Powers reported. He rationalized that there is more to be considered than just bestowing a universal storage OS to the masses. The key problem is "one of quality."
Despite his impassioned arguments that XIV was selected because of its support for generic, standards-based motherboards and chip sets, Powers insisted, "You can't just drop this technology onto anyone's box of disks."
By way of contrast, I had another chat last week with Krzysztof (Kristof) Franek, who is CEO and president of Open-E GmbH, a German-based firm with U.S. offices in Boston and Redwood City, CA. Established in 1998 as a sales representative and distributer of technology products, Franek now has 40 engineers in Poland and five more in the Ukraine who have done what IBM is currently unwilling to do. They have developed a storage operating system for the masses.
Available as software (you can download it from their Web site at Open-E.com to try it out) or as a USB chip that can plug onto any PCI motherboard's USB connector, you can make virtually any box of disks into a full-featured storage array. You can even select the attachment method you prefer: direct attach, iSCSI attach, or NAS. Just load a different USB key or software module.
Franek says that his product has already enjoyed over 20,000 downloads (in just a couple of months) and robust sales following the trial. His keys to success are simple: "Offer a high-quality product at a low price."
As a corollary, he offers that companies should listen to consumers and fill their feature/function requests to create the kinds of storage OS features that they need, and not to load them up with features they will never use but must nonetheless pay for.
From where we're standing, Open-E, and not IBM, is on the right path. Your comments are invited: email@example.com.
Jon William Toigo is chairman of The Data Management Institute, the CEO of data management consulting and research firm Toigo Partners International, as well as a contributing editor to Enterprise Systems and its Storage Strategies columnist. Mr. Toigo is the author of 14 books, including Disaster Recovery Planning, 3rd Edition, and The Holy Grail of Network Storage Management, both from Prentice Hall.