In-Depth
Unsolicited Advice for NetApp
Management shuffling is not what this company needs.
Trade publications have been filled recently with articles about NetApp (formerly, Network Appliance) -- the Sunnyvale, CA storage vendor that wrote the book on network-attached storage (NAS) almost two decades ago (albeit with considerable help in the form of technology assets acquired in a fire sale from distressed Silicon Valley neighbor Auspex).
In late August, there was a change of management at the company -- long-time CEO Dan Warmenhoven moved out of the driver's seat, and Tom Georgens, formerly president and COO, took the helm.
Speculation was rampant in the blogosphere and trade press that this move was predicated in part on the failure of NetApp to win a bidding war against EMC for the acquisition of de-duplication vendor Data Domain. This explanation appears to lack credibility, however. For starters, the price paid by EMC for Data Domain seems misaligned with the actual market potential for hardware-based de-duplication solutions going forward. Companies like CA are already providing very good de-dupe software engines for free as part of other software products, such as tape backup, which can be used with anyone's disk drives to serve the primary function that most companies have for de-dupe: waste management of backup files.
Wedding de-duplication software to a proprietary controller and a box of overpriced disks, regardless of the appeal of this approach to venture capital backers who prefer to wrap software in tin, seems well on its way to losing its luster. A lot of Data Domain boxes would need to be sold to recoup the $2.1 billion that EMC shelled out for the company.
In the end, EMC may have won the bid but lost any market share improvement they expected from the acquisition. By contrast, NetApp walked away with $57 million of Data Domain funds just for the privilege of being the jilted lover.
The last few years haven't been especially kind to Mr. Warmenhoven in terms of the value of product build and buy efforts initiated on his watch. In the first half of this decade, the company tried to spread its wings into the low and high ends of the storage market. In 2006, they introduced StoreVault to pursue the coveted SMB market that every enterprise storage vendor was chasing at the time -- a reflection of the already flat-lining sales of storage subsystems in the enterprise space. However, they discontinued the product in February 2009 after failing to make it work with Windows 2008 Server domains or developing support for jumbo frames in Ethernet that might have helped justify its high price tag. Most users ignored the company's offer to upgrade them to the mainstream NetApp product family and switched to other vendors.
The company's purchase of Spinnaker Networks in 2003 for $300 million was aimed at the storage clustering market, the one where supposedly massive storage repositories live. This move also failed to produce any sort of viable product for the company. (That said, Spinnaker-derived ONTAP GX technology seems to be on everyone's mind once again in conjunction with NetApp's announcements in the "cloud storage" space over the past couple of weeks -- see below.)
Similarly, acquisitions over the past few years of FC fabric management software maker Onaro, storage encryption vendor Decru, and data protection/data recovery vendor Topio all failed to extend significantly the company's reach outside of its "captive markets."
Meanwhile, cash reserves dwindled at NetApp to a reported $2.7 billion (well below competitors such as EMC), which helps explain why it did not try to outbid EMC for Data Domain. Moreover, the company that once boasted over 70 quarters of consecutive growth was lagging well behind its name-brand competition in Q1 sales figures in 2009.
In today's market, a financial snapshot alone is not sufficient to explain NetApp's change of management. All of these things considered together, however, the board may have decided that the time for new leadership had come.
After a round of musical chairs in the front office, the question always asked is "what will change?" Of course, Georgens isn't actually new to NetApp, having been directly involved with day-to-day management decision-making there for nearly five years. Even so, there are usually some new ideas or strategies that go along with a new boss. So far, we haven't seen or heard much that's new from their corporate PR machine.
Tapping the Cloud
On the heels of Georgens's installation, NetApp has been filling the airwaves with a less than compelling rant about cloud storage. Their spokespersons have been delivering the message to any trade press writer or analyst who will listen that the latest version of their NAS operating system, DATA ONTAP 8, is designed explicitly to facilitate cloud storage service providers. Their talking heads are saying that cloud storage requires seamless data motion between physical resources, a global name space, an always-on infrastructure, deep integration of virtualization technologies, increased automation, and chargeback features. ONTAP 8 blends ONTAP 7G features, designed to support multi-tenant virtual server environments, and ONTAP GX, the clustered NAS OS built at least conceptually on Spinnaker intellectual property.
NetApp is betting a lot on the purported appeal of cloud-based storage services, claiming that many of its customers and prospects are exploring the cloud option now and beginning to flirt with moving both packaged applications and application-test-and-development activities into cloud-based repositories -- the former, packaged applications, because it is increasingly easy to do, the latter, application test/dev, because of the rapidly changing resource requirements that characterize that activity.
They argue further that ONTAP 8 will support the need for "data mobility," a hazy term at best used alternatively to describe the movement of data storage connections between different virtual and physical platforms, the ability to add storage capacity, to move data without operational downtime, and to maintain current infrastructure components without disruption. As with most cloud storage terminology, "data mobility" sounds a lot like the functionality set that you can already get from storage virtualization software offered by companies such as FalconStor Software and DataCore Software, which can be used across any storage platform, including gear that is far less expensive than NetApp's.
As for the purported value of ONTAP 8 for improving data management, NetApp's talking points ignore hierarchical storage management (HSM), which moves less frequently accessed data to lower-cost storage on a routine basis, or CRON jobs and backup software that can copy data without disrupting the production environment or creating downtime. Does NetApp do these things better or with less expense? No one is making this claim.
The company is also emphasizing service level management functionality in its new cloud storage pitch, which is about the only thing that any of the cloud storage advocates are addressing with their wares that traditional storage purveyors all but ignored. I recently asked one cloud storage software company, Parascale, what was so special about their software stack that I couldn't get with storage virtualization today. As we worked through his model of cloud storage stack functions, the only feature that stood out was service level management, which is required to provide storage to multiple customers if you are in that line of business.
NetApp's cloud storage discussion hits this service management point hard, claiming that it will enable the segregation of storage users by role, enabling chargeback based on a more exact accounting of capacity usage. This may come back to bite them, however, since (1) companies have up to now shown only minimal interest in chargeback models and (2) since analyses of NetApp Filers, such as one presented last year at USENIX, showed that about 70 to 90 percent of the data stored on the Sunnyvale boxes that were assessed had not been touched in the previous year! All things considered, efforts to set up a chargeback service for NetApp storage will more likely than not produce a better argument for tape or optical storage deployment rather than any business value case for buying more Filers.
Truth be told, NetApp's ONTAP 8 discussions closely parallel the market speak of EMC around its "cloud storage" offering, Atmos -- which may be the point. Increasingly, NetApp seems inclined to follow, rather than lead, storage industry marketecture, especially that emanating from Hopkinton. Their purported rivalry with EMC has always struck me as a bit silly on its face, since NetApp's product set has always been more specialized than EMC's, and its NAS solutions have always outperformed anything that EMC could deliver with NFS/CIFS connectivity.
One of the biggest failures of the Warmenhoven regime was to frame enterprise data center storage as an "either/or" decision between the two vendors. NetApp specialized in files, EMC was all over the board, but focused mainly on block-level storage. Given NetApp's earlier hegemony in the NAS space, and given the fact that over half of any company's data is stored in the form of "unstructured" user files, I am hard pressed to understand why the company decided to confuse its value proposition with repeated references and comparisons to a block storage vendor.
In its manufactured contest with EMC, the corporate culture seemed to have changed in Sunnyvale. As an innovator in file storage appliances, Network Appliance was a vendor with a "mercantilist" mindset -- seeing the world as an ever-expanding market in which competitors only reinforced the validity of their network-attached file storage model. By contrast, EMC had a "Kameralist" mindset, perceiving the market as a "zero sum game" in which anyone else's success amounted to a lost revenue opportunity for Hopkinton. Today, it is getting very hard to see any difference between the companies: NetApp has become what it beheld.
A Four-Point Plan
My unsolicited advice to Mr. Georgens is as follows:
First, revisit both ONTAP and the other sacred architectural cows at NetApp, such as Write Anywhere File Layout (WAFL). Something in your brew of RAID, WAFL, and NVRAM spoofing is limiting the capacity that can be deployed behind your storage head. The press is increasingly filled with complaints from customers about your scaling limitations, which are being hit much more quickly now that more capacious drives are entering the market.
No, clustering appliances aren't the answer. No one prefers complicated clustering arrangements to much simpler capacity scaling approaches via storage virtualization or the horizontal scaling of physical arrays conceived as Lego building blocks. If NetApp's older technologies, like WAFL, stand in the way of change, it may be time to move on to a new architecture.
Second, focus on manageability. Configuration management, capacity provisioning, and status monitoring should not be the hassle today that they were ten years ago., yetcompared to Xiotech or even Overland SNAP, your management facility still challenges admins and requires a special training program. This needs to change and a Web services standards-based management approach is probably the right option to pursue.
Third, bring back Dave Hitz, who was responsible for most of the meaningful innovation done by your engineering department. Hitz is a straight shooter and a genuinely bright fellow. Since his departure from actual product development, after adding much-needed CIFS support to your products, critics argue that all changes to NetApp architecture have been incremental. In the words of one, "Now the company is left with a large customer base looking for real improvements in technology, and all they get is incrementalism and a kaleidescope of expensive software options." Assuming Dave wants the job, get him back into the thick of things. If he doesn't, you might want to spend some of your cash reserves recruiting somebody else with demonstrated smarts such as Steve Rogers at Overland (not that he would move from his beloved SNAP Server development effort).
Finally, as storage technology goes, NetApp solutions cost too much. That "middle class" buyer who has traditionally purchased your wares is now shopping for storage with a much reduced CAPEX budget. Today's tech buyers are also savvier about OPEX -- "soft costs" for things like software license and maintenance contract renewals. The ones I talk to are rethinking the wisdom of buying products that are rendered "end-of-life" by their vendor after only 18 months in the market and that have no resale value because of restrictions imposed on software license transfers.
One thing that NetApp has not perfected is the front-office sale: EMC's specialty. That means the company cannot succeed in selling low-value solutions by schmoozing buyers possessing low storage IQs the way that Hopkinton does. Your sale must be to the technically astute who need ammunition to justify the premium price you're are charging. Unless NetApp wants to see its gear traded in as part of a competitor's Cash for Clunkers program, the hard and soft costs of its products need to be re-evaluated and a better business value case, supported by facts, must be developed.
These four changes would be more important to your customers than all the cloud storage and cluster-speak that the company is peddling now.
Your comments are welcome: [email protected]