NetApp operations chief talks growth
Tom Georgens, executive vice president and general manager at NetApp -- also rumoured to be in the running as next CEO -- discusses what's driving its growth.
According to the latest IDC numbers, NetApp grew revenue 39.9% year over year. What's driving your growth relative to your competition?
Georgens: There are a couple of dimensions to this. We have a single architecture that serves NAS [network attached storage] and SAN [storage area network] protocols at the same time. The company's history has been as a NAS provider, but a third of our products out there are now serving blocks, some blocks alone and some with file data. This makes it easier to deploy the systems; customers don't have a different set of backup and disaster recovery products, and there's one set of administrative tools all the way through. It also gives NetApp tremendous leverage. We're pouring all our money into a single architecture. We have boxes with five or six drives and then a system we introduced in May that scales up to a 1,000 drives.This was NetApp's advantage at one time. But now it seems you have three very distinct product lines: the StoreVault SMB [small and midsized business] product; the enterprise line; and then the high-end OnTap GX product -- and all these run a different version of your core operating system.
Georgens: StoreVault is running Data OnTap, a version of Data OnTap that's the same as everybody else is running. But what we've done to go after that market effectively is to make the product easier to manage. So, there's a lot of functionality that's turned off, and admittedly, the StoreVault is like the Linksys to Cisco [Systems Inc.]'s enterprise products, but from an engineering and development perspective, there's still a tremendous amount of leverage.
Right, but I can't swap drives between a StoreVault and an FAS200 system. I have to buy a different box to get into the enterprise line of products. How about OnTap GX, that's a different OS [operating systems]?
Georgens: GX runs on the same hardware as 7G today. The WAFL file system, which is at the core of everything we do, is the same on both GX and 7G. GX is not like an HDS [Hitachi Data Systems Inc.] Lightning or Thunder, or EMC [Corp.] Symmetrix or Clariion, it's an underlying OS. But all the things you'll see on top, our flex products, our snap products, will all be common. Clariion and Symmetrix are totally different hardware platforms, different operating systems -- they have different replication and management products. That's not going to be the case with GX. You'll be able to replicate between them, migrate between them. They will interoperate and coexist with OnTap 7G systems as all the premium features will be common.How long before all your current features are supported in GX?
It's going to be a multiyear thing. We've learned that it takes customers a long time to make transitions. Even with 7G and really cool features like thin provisioning and flexible clones, a lot of our customers in heavy production with 6.x, are taking a year or two years to transition. So you'll hear a lot from us about coexistence and migration, as opposed to 'OK, everything is on GX and you need to cut over.' That's unrealistic in terms of customer migration. Next year, you'll start to see integration of our premium features into GX to bring GX as part of our mainstream OS to serve not only our high-performance technical customers but also our business customers and our customers in mainstream production.So if I want the next-generation Snapshot capability, I will be forced to migrate to GX?
I wouldn't say forced, over time there will be, on a market-by-market or workload-by-workload basis, a clear advantage of one versus the other. But they're going to coexist for a long time. We expect some customers will be running some applications on GX and others on 7G, all managed from the same user interface and all using the same snapshotting and replication tools. We're developing a layer of abstraction above the operating systems so that the premium features perform independently from the operating system they are running on so that we can decouple those two developments. The more we're investing in a GX stack and a 7G stack, the more we're splitting our resources. We clearly don't want to do that … Admittedly, there's a burden on the user interface. GX has the ability to scale out and to manage many machines as if they were one, while 7G is the single box construct. So it's key that the user interface contemplates them both, but the features themselves will operate abstractly from the operating system. Still, the user interface has to manage the complexity of both different types of systems and present an easy migration solution to the customer. The user interface has a fair amount of duplicate work, no doubt about that.How long before you end-of-life 7G?
Georgens: Well, since we introduced 7G, we've already done two major releases, and we've got several more major releases along the way on our engineering roadmap that are staffed and heavily underway. But we don't want to be doing stranded R&D [research and development] for an old operating system that isn't going to live or stuff that we're going to do for a new operating [system] that's incompatible; we're working really hard on the migration strategy and the compatibility so that we don't waste a lot of engineering on stuff that has no future.We heard the transition from 6G to 7G has been difficult for some users. Why is that?
Georgens: To take the full functionality of 7G, there are some data format changes, and I'll admit that's not the easiest transition, but customers are doing it. The features are compelling, and they are making the conversion. It certainly isn't hurting us, but there's a newfound sensitivity for customer migration watching some of our bigger customers transition from 6[G] to 7[G].How come GX is so expensive?
Georgens: It's expensive because these are really, really big, high-performance computing configurations. GX is a clustered system, so it's all the storage that goes with it. If it was a two-node cluster, the price would come down considerably. The price point would be the same if the customer were buying a whole bunch of individual boxes versus a GX system. There's nothing inherently expensive about GX.Switching topics, what's your largest iSCSI deployment?
Georgens: I can't really talk about individual customers, but the most meaningful application is Microsoft Exchange on iSCSI … but iSCSI is not constrained to the low end and the small ones; in fact, we had a customer at our analyst day that was deploying, not Exchange but iSCSI for Oracle, and that was working out well for him. Still, it's primarily a Windows product, but we see some iSCSI and AIX and iSCSI in Linux, but some of the other operating systems either don't support it or don't support it very actively.What do you expect to happen when Microsoft's iSCSI target software hits the market?
Georgens: The Stringbean stuff … I think in the enterprise space that I'm responsible for, we've got plenty of ways to fight that off in terms of breadth of functionality, maturity and scalability. On the lower end in the StoreVault space, this'll be something else that HP [Hewlett-Packard Co.] and Dell [Inc.] and everybody else that depends on Microsoft down there for their offerings can add to their portfolio. It will be something we need to contend with. Time will tell how mature and fully featured and well integrated it's going to be. In the enterprise, it won't be a big deal, but for StoreVault, we'll line up against it everyday.In the CIFS world and iSCSI world, there are some cost advantages for us as the competition are using the Microsoft tools, but since we have our own, that gives us some flexibility both in terms of a differentiated feature set and how the product is priced and positioned.
We hear the FAS3000 is taking share in the midrange. Is that a price war in that space?
Georgens: The 3000 is the middle of our product line, and the middle of the product line is usually the most cost-effective piece of hardware. Usually the high end is optimised for performance and the low end for cost. So it stands to reason that the midrange of the product line would be the highest volume. But there's another dimension to this. If you talk to customers, it's not just primary data that's the source of major headaches (how do I serve my database, how do I keep my critical applications running) there's also what to do for business continuance, for backup, for compliance. I want copies of my data for test and development, I need copies of my data for decision support …If you look at the IDC numbers, we're driving very high terabyte growth. That's because we're battling hard on the primary [storage], but I think we're making it very easy for users to manage all these secondary copies of data with one integrated offering and single data format. After we sell the primary, we're pulling a tremendous amount of secondary storage with it.
NetApp has also moved into the No. 2 position for data replication software, and we've achieved that on a much smaller footprint than our competitors. It means that the connect rate of our replication software is higher than that of many of our competitors. People are using SnapMirror for more than just business continuance, they're using it for disk-to-disk backup, for archival and compliance. A lot of our terabyte growth is coming from customers deploying more sophisticated data management strategies that not only talk about primary storage but integrate business continuance and backup.
Beyond the usual data growth statistics, what applications do you see driving storage growth?
Georgens: We're seeing an acceleration of customers that can no longer meet their recovery times using tape. So we're seeing more and more customers using some type of disk methodology in their backup strategy, primarily to reduce their recovery time … The industry in general is very primary-storage centric, but a lot of the customer headaches and lot of the explosion of incremental data is coming in the secondary storage space.Then there are these Web 2.0 applications, like MySpace and YouTube and iTunes, clearly these are going to be a driver. As they mature, they will drive different kinds of data expectations and we're looking at that.
From NetApp specifically, you'll see us integrating as many of our premium features as close as possible to the application as we can. We have tools that can kick off backups and business continuity directly from Oracle or recover a single mailbox directly from Microsoft Exchange. Basically, we're bringing the storage tighter to the application, eliminating the purchase of some type of middleware or months of professional services or a manual process that's not going to pass 404-compliance, to close the gap between the applications and the storage. And [that gap] is bigger than we in the storage industry like to admit it is. We want to enable people to control their data directly from the applications themselves much more than they can do today.
It's no secret you're being groomed to follow Dan Warmenhoven as the next CEO at NetApp. Are you up for that job?
Georgens: [Laughing] Who says that? I have no real answer to that. I am excited about being here. It's a great company. But …You wouldn't say no?
Georgens: I wouldn't say no. But Dan and Tom Mendoza and Dave Hitz and James Lau are still very actively involved in the company and are adding far too much value for me to engage in that conversation.