A year ago around this time I was in Las Vegas seeking answers about what VMware had up its sleeve for Act 2. The answer, then as it is now, was simple: cloud. Not that they had any intention of abandoning their overwhelming command of the virtualization markets, of course. But with their current position not indefinitely sustainable in the face of open source hypervisor commoditization and the long expected arrival of Microsoft’s Hyper-V, it was imperative that VMware adapt and identify new opportunities for growth. And with Microsoft alumni Paul Maritz at the helm, it was only natural that they would seek a platform opportunity.
This year in San Francisco, then, I was looking for some feedback on their execution. Their ambitions were admirable, to be sure, but ambition is just another way of saying you’re not there yet. So herewith a few observations on their progress in the form of a Q&A, but before that a quick caveat.
VMware, like one or two other vendors, requires that analysts sign an NDA prior to attending the show. The stated reason for this is to permit a more open conversation, and it is certainly true that its executives were very candid about their thoughts with the assembled group. That said, NDA’s are obviously a significant impediment to subsequent public discussions and analysis. Over the course of an entire day’s meetings, it is very difficult to track the public or non-public status of each bit of information captured, as you might imagine. Particularly since the presentation materials used were not made immediately available on the flash cards handed out.
Having signed the NDA, however, and choosing as an organization to observe the spirit as well as the letter of such laws, I am obligated – legally and ethically – to respect the confidences requested. This means that the majority of the information learned – much of which is probably publicly disclosable – will not be discussed here, because it is impossible to efficiently determine which comments are public and which were considered private.
Therefore the bulk of this post is either high level observations that cannot be considered confidential, or known to be public materials. Much of the good, in depth content, then, will not be discussed here. Such is the cost of full NDA analyst days.
Q: Before we begin, anything to disclose?
A: Sure. VMware is not a RedMonk client, but several of its competitors including Canonical, Microsoft and Red Hat are.
Q: On a high level, how would you describe VMware’s current transition from virtualization behemoth to cloud provider?
A: VMware’s position is not too different from Google’s or Microsoft’s, on a high level. They were enormously successful in what became a strategic market and have extracted significant revenue from that success. Like Google – so successful with search, VMware is seeking its next big marketplace success – its Microsoft Office, if you will. This is an imperative for the firm not simply because it’s good business, but because VMware’s marketshare – like Microsoft’s operating system share – is likely to erode over time. Not because of any simple product or business model limitations; the firm continues to innovate both around their core virtualization products as well into newer, greenfield opportunities. VMware’s marketshare is likely to decline rather because of certain marketplace realities. First, free is difficult to compete with, and the aggressive injection and maturation of open source alternatives in the virtualization and management areas would inevitably be felt. Perhaps more importantly, Microsoft – with its global ubiquity – is in the process of introducing competing technology at the lowest pricepoint they can support without becoming the target of further antitrust investigations.
VMware, perhaps more than anyone, knows all of the above. Which is why we saw the public disclosure last year of their future plans. The cloud’s timing could not have been more fortuitous for the vendor, in that it promised a potentially massive new market to which virtualization was not only an option, but a necessary foundation. Even better, it was sure – for compliance and security reasons, to say nothing of traditional enterprise caution – to be adopted at a slow enough pace to give VMware time to pivot and adapt its software portfolio to the needs of the cloud.
Which is what we saw at this year’s VMworld.
Q: How so?
A: Well, one of the things that VMware spent a lot of time talking about – both last year and this year – was the channel, and more specifically, the hosting service provider portions of that channel. The attraction of the service provider market is obvious: they were slow to wake up to the threats posed by cloud computing market entrants such as Amazon, and to add insult to injury will be challenge to compete effectively with the economies of scale that such competitors can bring to bear. What they are desperately seeking, then, are technical solutions that will allow them the ability to offer cloud-like services on their existing infrastructure without the cost of having to build them in house. Which they probably couldn’t do anyway.
Given that many hosts – at least the mid-size and smaller versions, anyway – are built on top of open source software, in an effort to keep their license costs to a minimum, it may yet be that open source projects like Eucalyptus will be the first option they turn to for their cloud solutions.
But just as products like VMware Workstation have a market not because they necessarily do things open source can’t, but they make them easier, VMware is likely to compete aggressively on the basis of its Microsoft-like integrated innovation. The interesting thing about VMworld this year, however, was that it was more than just talk: we’re beginning to see execution here.
Q: With which partners?
A: The current vCloud Express page lists as its partners Bluelock, Hosting.com, Logica, MelbourneIT and Terremark, but they also trotted out speakers from AT&T, Savvis and Verizon during an analyst cloud session on Tuesday morning. The importance of that list, to me, are the different pricepoints.
Q: How do the pricepoints factor in?
A: Make no mistake about it, one of the primary assets VMware has in its battle to win the cloud is its enterprise acceptance. While most cloud startups – and even Amazon, to a certain extent, virtual private clouds or no – will have to answer and reanswer questions about their technology, their long term viability and the quality of their support organization, VMware is a known – and for the most part, trusted – entity with most of the Fortune 1000. They don’t play the same strategic role within each, of course, but VMware is a name not too far behind Cisco, HP, IBM, Microsoft, Oracle and SAP when it comes to enterprise brand acceptance.
That said, the mistake that many enterprise cloud players continue to make is not making their technologies available to the common man: the individual developer, the startup, the curious enterprise developer. Too many would be cloud offerings start with: call us for details. If VMware’s sole partners in this endeavor were AT&T and Verizon, you could probably expect the same.
By partnering with the likes of Terremark, however, VMware ensured that their technologies are sufficiently adoptable that I was able to create an account using my RedMonk credit card before the first speaker at the cloud session was finished talking (see above). That’s enormously important, in my view, because it makes the technology relevant and visible to everyone from the CIO down to the guy in the trenches just trying to get things done.
Q: What are the pricepoints?
A: They’ll vary by provider, obviously, but the Terremark account I signed up for is relatively low cost. At $0.037/hour (including the $0.002 cost for a 4 GB system disk), the cheapest Terremark instance undercuts the smallest Amazon EC2 option by $.063/hr, which for the sake of context is a $1.51 difference per day and $10.58 per week. Not much when you’re talking about one machine, but scale that out to hundreds or thousands of instances and it would add up. In Amazon’s defense, it should be noted that a.) its small instance has better than three times the memory relative to the Terremark option (1.7 GB to .5), and b.) the available disk space is much greater.
Suffice it to say that VMware is going to have partners that are price competitive with the likes of Amazon, as well as higher end partners for the virtualization provider’s traditional higher end clientele.
Q: Let’s go back to the open source question for a minute? Are there questions about openness in the cloud?
A: Absolutely. More so, probably, then there are regarding on premise infrastructure. Because the cloud acts to collapse many previously separate technology choices – hardware platform, storage subsystem, networking gear, operating system, programming language, database and so on – the potential for lock-in is exceedingly high. Which has prompted a commensurate level of interest in quote unquote open cloud solutions, though what that term means, precisely, is often unclear.
When asked the question, VMware’s Bill Shelton pointed to, among other things, VMware’s vCloud API which – besides being royalty free and extensible – has been submitted to the DMTF. So while the source code for VMware’s products will remain closed, the interface will open, documented and potentially a standard. Whether that will be enough for customers remains to be seen. Whether it should be enough from customers depends largely on your philosophy.
In theory, open APIs are excellent – provided that they’re comprehensive enough functionally – as they facilitate interoperability and thereby, to some extent, openness. The difficulty from a customer perspective is the volume of differing standards; just look at the list that the CohesiveFT gang has accumulated. So while it’s a positive that VMware has chosen to document and standarize their API, that’s still one more interface for would-be customers to learn and adapt to.
Q: What about the operating system? Did we learn anything new about its importance at the show?
A: Nothing more than what we already knew: that for VMware, the operating system is essentially an afterthought. Maritz spoke publicly, as he has in years past, about “cutting the tentacles” that link the operating system to the hardware, thereby offering you greater freedom of choice for both hardware and operating system. While VMware seems to have its focus turned away from the virtual appliance notion at the moment – a notion of which I have been skeptical – there is little question that VMware’s mission is to disintermediate the operating system from a foundational perspective. That this will be easier to do in a cloud environment which is intrinsically less operating system focused is perhaps one more reason for the focus in that area.
Q: What about SpringSource? Was there talk of that acquisition?
A: Far more than I’d expected, actually. Besides the obligatory focus on the use of SpringSource as a jump point for PaaS offerings, the part that interested me – that I’m pretty sure I can talk about – was the fact that Maritz clearly understands that there’s a fairly fundamental development shift underway. Frameworks are playing an increasingly important role in the ever more fragmented development arena, and in the Java ecosystem that dominates, along with .NET, the CIO agenda at VMware’s traditional customerbase, there is no more popular framework than Spring.
Q: What do you think is missing from VMware’s cloud discussions thus far?
A: There are a couple of areas that I think could benefit from some attention, but the one that I discussed with VMware was migration. This is a common problem in the industry, frankly; too often the described benefits of cloud computing are only available to greenfield workloads. The analogy I used was Google Apps. The mail application is performant, interoperable and has a good mobile story. But until the introduction of IMAP import, you’d essentially be leaving all of your old mail content behind. With that feature, however, you can trigger an import and be transparently up and running on the new platform with all of your data seamlessly transferred along with you.
This apples to oranges, of course: migrating between mail servers is slightly more trivial than between on premise and off premise infrastructure options. But vCloud Express would benefit tremendously from an IMAP import like feature – one that could seamlessly import and export runtimes to expedite the process of getting from Point A (on prem) to Point B (the cloud).
Q: Any other questions? How does this compare to the just announced Deltacloud, for example?
A: I just got back from California yesterday, so I’m goint to take a day or two before I get to that one. Any other questions on VMworld, drop me a line.