I had to take a break from my sequence of posts recapping this event on March 10th, what with Cisco's UCS launch, and the whole IBM/Sun thing.
But I'm back to the task at hand.
Why?
So many important and interesting concepts came out at this event that I'm committed to keep working through the sequence -- it was a treasure trove of advanced IT thinking, and deserves wider exposure.
The last post covered the first half Paul Maritz' take on VMware's strategy. So much was said that I had no hope doing it all justice in a single post -- even a very long one.
So, let's get back into the flow from Paul Maritz.
A Quick Reprise
Early in his presentation, Paul put up an "anchor slide" with three core customer benefits: efficiency, control and choice.
And he positioned the three major VMware initiatives: vSphere (shown as VDC-OS on the slide), vClient and vCloud against that backdrop.
In the previous post, we did justice to vSphere (formerly ESX) -- now on to vCloud and vClient.
The Cloud As Service
Earlier on in the presentation, Paul made the eloquent observation that "cloud" really represented two related concepts -- one, how computing architectures should be built, and the other for how they're consumed.
It certainly would make it easier to have a productive discussion :-) You know, I've seen people get so wrapped up in cloud-y religious debates over this simple duality, that I'm hoping that we can partition the industry argument into "cloud as architecture" and "cloud as service".
The vSphere initiative speaks directly to the former, the vCloud initiative to the latter.
The key concept here is enabling choice -- where to use entirely internal resources, as an example. Or entirely external resources. Or any dynamic mix in between. Or having access to an open marketplace of compatible and federated service providers.
Many of us believe that the current "cloud providers" are basically one-offs -- the workloads aren't easily portable, there's no hope of federation internally/externally -- let alone service providers federating between themselves.
If you think about how phone companies and shipping networks and power grids work today, you'll see an open marketplace of compatible infrastructure providers who freely federate to meet customer requirements. We'd like to see this in IT infrastructure as well.
Simply put: getting the "right half" of his slide enabled is what vCloud is all about. And, as you'll see, VMware will be getting some help from EMC and other vendors in this regard.
Surprising vCloud Momentum
Last year, when VMware first announced their vCloud initiative, they pointed to something like 100+ service prooviders and outsourcers who signed up.
Well, since then the number as grown to over 400. I've seen the list of names -- it's basically a who's who of the industry.
Personally, I've been in a few high-level meetings with some of these outfits on this very topic. They're deadly serious -- this isn't some fan club roster we're talking about.
And these are people who know how to stand up carrier-class infrastructure and charge for services delivered. They understand things like availability and reliability and predictable behavior and real-time fault isolation and change control and security. Frankly, they stand in stark contrast to many of the names that get bandied about as "cloud" players.
Paul makes the stunningly obvious point that any decent consortium of compatible service providers will dwarf the current "cloud club" by many, many orders of magnitude -- making Amazon, Microsoft -- even Google -- look quite small and quaint by comparison.
These service providers understand how to make massive infrastructure investments at a global scale.
And they have a whole lot of data centers and bandwidth in the ground today.
Do not think they will not be a dominating force in the move to cloud.
Real World Examples
It's tempting to sit back and say -- boy, this will all be cool someday, let's just wait -- but Paul wanted to make the emphatic point that there were serious examples of vCloud-based offerings in the market today -- with many more to come.
And, like any other open marketplace, specialization and differentiation was already evident among the current players -- and more was to be expected.
Using a standard VMware-based stack, some had chosen to focus on availability, or cost, or security, or specific verticals, or geographies -- a very rich ecosystem was forming.
The idea of dozens -- or hundreds -- of compatible service providers, each willing to run your virtual machines as you want -- well -- that's a rather breathtaking vision, if you fully consider the ramifications.
I've included some of the slides as small thumbnails you can click on if you want details on the companies he highlighted.
On To vClient!
Paul then introduced the key idea behind vClient -- that IT should be in the business of provisioning users and application experiences, and not provisioning devices.
From the user perspective, the user experience (and its information) should follow the device -- and leverage any native capabilities that might be available in the device.
As a multi-device kind of person, I want to live in that world sooner than later!
He then showed how two basic technologies -- server-hosted desktop virtualization aka VDI -- and a newer model (client hypervisor aka CVP) would cover the broad range of required user experiences using a single framework.
Side note: gee, the MVP (mobile platform) demo shown at VMworld was so darn cool -- wonder why he didn't draw the picture out even further?
The result is that IT can begin to think in terms of "desktop as a service"; largely independent of device, location, etc.
Like the private cloud model before, the idea is to facilitate choice for IT and users: continue to provide desktops as before, let the user choose the device, or any mix of the above.
As many people know, the entire end-to-end suite will be known as VMware View, it will include provisioning and management tools,, and it will be rolled out during 2009.
Personal note: I think the desktop virtualization market will go to a vendor that can (a) provide a consistent user experience across multiple client models: thin, thick, mobile, (b) has the management tools to make it simple for IT, and (c) is compatible with the server-based application infrastructure.
When put in those terms, it's hard to bet against VMware.
EMC and VMware
Perhaps the headline of the day was Joe Tucci telling the street that EMC wasn't considering divesting itself from VMware anytime soon. He wanted to put all speculation to rest -- once and for all.
Joe presented his reasons as to why he thought the way he did, but Paul also thought it important to share his views as the CEO of VMware.
He put up two slides to illustrate his perspective -- one at a business level, the other at a technology integration level.
Both Joe and Paul commented -- repeatedly -- that any work and collaboration between the two companies was through mutual advantage, and not through corporate edict.
Both EMC and VMware were free to partner with others, and did often.
Paul reinforced this point that he had the privilege and the responsibility to run VMware as a free-standing company, and Joe along with the rest of EMC was living up to EMC's part of the bargain, e.g. hands off.
Paul also mentioned that the techincal collaboration between the two companies was extensive, and was widening.
It's important to point out that this technical collaberation was open to all vendors, including EMC's direct competitors.
Since VMware's strategic play was in large data centers and service providers, EMC brought extensive credentials to the table as a partner: consulting, solutions engineering as well as certification and test at scale.
Now, technically speaking, other VMware partners could do the same, but I think EMC has invested far more aggressively in the VMware partnership than most.
Don't underestimate the value of stable shareholding either. Paul doesn't have to worry too much about who's buying his stock, keeping institutional investors happy, or any of the other joys of being a large cap software company with a diverse set of shareholder interests -- freeing him and his team to focus exclusively on the opportunity at hand.
And then, thankfully, Paul spent a minute highlighting where EMC was contributing to some of VMware's initiatives.
For the vSphere initiative, he pointed at storage APIs (both development of these APIs, as well as direct support of them) and newer security and DLP APIs (again, help in defining the APIs followed by direct product support).
He also mentioned -- very briefly -- the work the two companies are doing to harness the newer intelligence of storage into large VMware environments, but didn't give a lot of detail to this specific audience.
For vClient, Paul thought that thin provisioning of storage was important. I'd add the security stuff and service delivery management here as well, if it was my slides. And, of course, our solutions and consulting work in this arena.
And then for vCloud -- not only was their direct applicability of EMC's technologies in storage, resource management and security, but more was possible as well.
And Now, Back To Joe Tucci
I have to admit, I was a bit breathless after Paul was done. I mean, really, to cover all of this in a very short period of time, and to fully understand all the implications -- well, that was a lot.
For the next post in this sequence, I'll cover how Joe put it all together in a go-forward strategy.
Hello Chuck,
It was useful to know the thought processes. I am however trying to understand how the "last piece of this jigsaw", which to me is the "last mile access to home or office" will be addressed. I just did some elementary maths on the bandwidth requirements: Assuming that a small office with just 20 people, use a cloud (with local thin clients) for their normal work. Also assuming that that most use windows and remote desktop into their Virtual desktops within the "cloud". For a seamless user experience, the link minimally has to be 1 Gige (100 MBps is a bit of a drag). That effectively means that each of the 20 users need a Gige connection on their machine/thin client. In effect, the office needs about 20Gige of external bandwidth into the cloud..and that just covers 20 users! Closer to the head, the bandwidth handling demands increase significantly more because of the accompanied aggregation.....this by itself seems to call for phenomenal investment in the last mile path, unless compression technologies improve vastly. Am I missing something here or are we looking at something really far into the future?
regards
sudhir.brahma@gmail.com
Posted by: Sudhir Brahma | March 30, 2009 at 11:35 AM
Hi -- you are assuming a thin client model only.
The CVP model says that the client has a full hypervisor and virtualized desktop machine running locally, and can operate autonomously if needed.
Changes get pushed/pulled as needed.
The bandwidth requirements in such a scenario drop dramatically.
Posted by: Chuck Hollis | March 30, 2009 at 04:51 PM