Analysis: HP Unveils NeoView
What makes HP think it can succeed where so many other vendors haven’t? A unique mix of software, services, and fault-tolerant servers, for starters.
- By Stephen Swoyer
- May 9, 2007
Get ready to cue up the Einleitung from Also Sprach Zarathustra: late last month, Hewlett-Packard Co. officially unveiled NeoView, its long-awaited data warehouse (DW) appliance—or quasi-appliance—offering.
NeoView is a quasi-appliance for the same reason that Teradata—which is currently a quasi-independent company (it’s still a division of NCR Corp.)—is also a quasi-appliance: it’s based on a combined hardware and software platform, just like an appliance, but it also bundles installation and configuration services, business continuity planning and capacity management services, and other amenities, too. That’s a market segment (and a value proposition) which Teradata seems to have pretty much all to itself.
So what makes HP think it can succeed where other vendors (including past luminaries such as Red Brick) have gone under?
HP’s thinking goes something like this: You don’t necessarily need a virtuoso like Buddy Rich to drum at your daughter’s wedding—although, certainly, you’d have Rich if you could afford him. (And if he were still alive, natch.) But, really, almost any combination drummer/bandleader would do—provided he or she can keep time. But what if you could get a slightly less prominent drummer—not a Buddy Rich, per se, but a Philly Joe Jones, or a Max Roach, or a Tony Williams (you’d better forget about Keith Moon; your insurance underwriter would never agree to it)—to play the gig at a tempting price?
That, in a nutshell, is NeoView’s value proposition: it doesn’t purport to play in the extreme high-end of the DW market segment—the virtuoso space in which HP likes to pigeonhole Teradata—but it could, officials contend.
Moreover, claims NeoView marketing manager Vickie Farrell, HP can plausibly go lower and wider and cheaper than Teradata, thanks to its expertise in developing and manufacturing a wide range of server hardware, and thanks to its enormous professional services arm—which it augmented earlier this year by acquiring BI and DW services specialist KnightsBridge Solutions.
"[Teradata’s] target market is [as an] enterprise data warehouse for very large customers. I don’t think we’re as niche, as high up into that, as Teradata is," says Farrell. "Yes, we have very similar architectures, and, yes, Teradata has done a great job proving that a shared-nothing, MPP [massively parallel processing] architecture is the best architecture for doing data warehousing. But we’re different."
HP has a shared-nothing, MPP architecture of its own --NonStop, which it inherited from the former Compaq Computer Corp. (which in turn acquired it from the former Tandem Corp.). Tandem, along with IBM Corp., Stratus Technologies, and Unisys Corp., was a respected provider of continuously available (e.g., fault-tolerant) and highly scalable computer resources.
In this respect, says Farrell, MPP’s combination of fault-tolerance and scalability (via clustering) is tailor-made for the dicey practice of data warehouse capacity planning. "It’s really hard to do capacity planning for data warehousing. It’s not as straightforward as OLTP. That’s one of the places where IT organizations get in trouble—trying to anticipate the needs," she comments. "One of the nice things about shared-nothing is that it’s pretty predictable; if you double the load and you double the system, you will keep the performance at the same level. And MPP is obviously very, very important when you’re trying to handle many queries simultaneously."
In other respects, Farrell argues, NeoView and NonStop go Teradata one better. "We are taking other things from NonStop, like the fact that the [NonStop SQL] database and the operating system are tightly integrated—like AS/400-type integration. You don’t even see an operating system [in NonStop]; you can’t even log on at a root level. There’s no system administration involved at all," she indicates. "The other is fault tolerance. We’re taking that built-in fault tolerance from NonStop, because we think it’s going to be a crucial differentiator. We’re starting to hear more and more from customers how important fault tolerance is. It’s an evolution, of course. It’s not like one day you didn’t need it and now you do. But in some of the places where we’ve gone in and done comparison benchmarking against Teradata, the customers are pretty pleased with the Fault tolerant capabilities."
HP, like Teradata, tends to downplay the competitive threat posed by the data warehouse appliance. Farrell, for example, says that NeoView and Teradata differ from DW appliances, which she says are typically deployed as data marts—or, similarly, as a means to address highly specific pain points.
"What we’re seeing in the market today is that people who implemented their first round of data warehouses are kind of running out of steam now—certain queries don’t run fast enough; they can’t support enough users; they’re having to put in more [hardware] to kind of handle the load," she says. "Customers are looking for a solution to this, and the appliances are an attractive alternative. But they really don’t expand to be an enterprise platform. People put it in for the quick fix, but they really can’t expand to leverage it. What we’re saying is, the NeoView is in the same price range, so use it. It does the same kind of table scanning, and because of all of the processors, it attacks the problem in the same way. But you’re not limited to just fixing the performance of those certain queries. You have something that’s flexible and that can grow as your demand grows."
There’s a further wrinkle here, too, says Farrell: you can say what you want about Teradata—and DW appliance vendors have had plenty to say on this issue —but its business model has been validated by sustained profitability and growth. In a sense, she concedes, NeoView—which HP aims squarely at the enterprise data warehouse (EDW) segment—also helps validate that business model. But for all of the attention the appliance model has garnered, Farrell points out, it has yet to demonstrate its viability as a profitable product segment.
"Netezza has been around for around four years now. Yes, they have 100 customers, but they’ve also done it by not charging a lot of money. Now the question is, can that price point profitably sustain a company? It hasn’t thus far, so perhaps price does need to be a little bit more," she says. "Now our price is a little bit more, and for that, you get an awful lot of services."
And services, Farrell contends, are what help separate the EDW leaders from the EDW leader-pretenders. "With six years at Teradata, I came to understand how much the customers appreciated all that Teradata provided. If I [were] a customer and I had a problem, I only had one number to call: Teradata. If the problem is with the BI tools, if it’s with ETL—it doesn’t matter. We think that a lot of customers out there are looking for that—and we go a lot further than Teradata in providing that. If anything goes wrong—with your hardware, software, whatever—you call HP."
HP billed last month’s announcement as NeoView’s official "launch" event. But what of HP’s abortive NeoView launch late last year? That, Farrell says, was more along the lines of a pre-release release event.
"[NeoView] was released in October to a limited sales audience of existing HP customers. In fact, we didn’t really have the full sales force in place at the time. Now we do have that in place, so what we did last [month] was really the formal launch and public announcement of the product really for the first time," she explains. "In that time, we were selling to some target customers. We were also doing a lot of testing of the product at customer sites as well as within in our own IT organization. They’re consolidating a lot of data marts and building an enterprise data warehouse, which is really a good testbed for this product and also a good showcase for what it can do."
Something else changed between then and now, too: HP picked up more than just a patina of BI and DW credibility, thanks to its acquisition of Knightsbridge. If potential customers had any doubt about HP’s BI and DW worthiness when it first unveiled NeoView, the acquisition of Knightsbridge should have helped put them at ease, Farrell argues.
"Knightsbridge has been around for 12 years now. I know how good they are, and they got to be as good as they are by working with every product out there. They’re objective. They’ve been very objective, and so one of the questions that people are asking them is that now that you’re part of HP are you always going to be recommending NeoView, and the answer is absolutely not," she concludes. "They do not want to lose that objectivity. They do not see themselves as selling NeoView. We’ll sell NeoView and have them come in to do the consulting, and, yes, we’ll get first dibs on having them come in and do the consulting, but they still will do the heterogeneous environments, too."
If nothing else, says Mark Madsen, a veteran data warehousing architect and a principal with consultancy Third Nature Inc., HP’s NeoView play will make things interesting. Madsen says he isn’t convinced that NeoView’s all-that-and-a-boatload-of-services pitch can stand up to the cut-throat economics—what Abraham Lincoln might have called the irreducible arithmetic—of the appliance wave. "HP seems to be focusing on DW and BI again, and their rise is corresponding with Sun's decline. NeoVIew is a different beast though, since Oracle and DB2 won't be running on it. Not sure how much traction they'll get given that the price-point doesn't seem as compelling in the DW market as something like a Netezza or Datallegro," he comments. "I'm just not sure how willing IT organizations are to give up the concept of a standard database for a variety of black boxes with a SQL pipe going into them."
At the same time, Madsen allows, market trends do change. Perhaps customers are ready for a lower-cost alternative to Teradata that’s priced at a premium relative to appliances, too. "Maybe the market is beginning to shift, since we really shouldn't care what's behind the SQL interface, only that it's cost-effective, performs well and is easier to manage than the current database-operating system mix we have now."