Sometimes when you hear about a concept, it grows in your mind until you finally get a full appreciation for its strategic significance.
I seem to remember reading that when Einstein offered up the notion of the relationship between energy and matter, it took a while its significance to become widely obvious.
OK, we're not talking about the basic structure of the cosmos here, but -- in its own way -- the notion of vClouds as presented by VMware this week may have its own underappreciated signficance.
It may end up being the way most businesses first take advantage of the cloud.
And, eventually, perhaps define how most business users of IT come to think about "cloud".
Read All About It
If you missed the announcement of the vCloud initiative, and the corresponding program to recruit service providers, please go take a moment to have a quick read.
Not surprisingly, the industry press tended to characterize this as yet another technology vendor jumping on the cloud bandwagon. That's fair, given all the interest in the topic.
But to simply think this is just like everything else you've heard about would be to miss something pretty significant, in my opinion.
And here's why ...
Between Two Extremes
One way of characterizing the transition to cloud is think in terms of minimizing disruption, and creating easy options for people to consider.
And, when I look at most of the current "cloud thinking" that is directed at corporate use of IT, I see the potential of a very painful journey for all participants.
Not only are we talking entirely new operational models, but -- in many cases -- we're talking entirely new applications as well. And asking an IT organization to do both at the same time -- well, that's asking a lot, isn't it?
Now, it all looks much better, IMHO. We're potentially talking a far smoother transition for all involved.
The Architectural Challenge Of Multi-Tenancy
A key concept in enterprise clouds is the notion of multi-tenancy. To reach peak efficiencies regarding resource utilization, you have to essentially build an environment from the ground up that assumes that you'll have thousands (or millions!) of distinct logical entities running the exact same instance of the application.
Put another way, you can't easily take an application designed to run in a corporate environment, and simply decide to run it in a multi-tenancy environment. Not only will you be missing all manner of functionality to "manage the tenants", it'll be horribly inefficient from an infrastructure and operational perspective.
One example is backup. All sorts of products exist for people to back up their desktops and laptops themselves (EMC's Retrospect comes to mind). But it's a different proposition entirely to offer that same functionality as a cost-effective cloud service (EMC's Mozy comes to mind).
Mozy is not millions of copies of Retrospect running: it's a purpose built environment that's designed to do one thing and one thing well. It's a ground-up exercise today.
Now, consider a moment the vast diversity of business applications that people use to run their businesses today.
How many different customized and tailored code stacks are there out there? And how many are amenable to complete re-architecture -- and user experience -- to run efficiently in a multi-tenancy environment?
And will people be willing to migrate entirely from a tried-and-true comfortable set of business applications that they've used for years, to something entirely new -- and very different?
Sure, there will always be customers in the marketplace looking for something new, better, etc. -- but when compared to the vast legacy installed base of business applications, this "willing to try something new" crowd is a small sliver of the total opportunity.
How do we get everyone else to the cloud?
The Beauty Of The Virtual Cloud
We all know that virtualization technologies such as VMware encapsulate existing applications, and isolate them from the physical hardware. Paul Maritz spoke eloquently about this in his keynote this week, referring to the "deconstruction of the operating system".
Pushing the concept a little further, virtualization also can deconstruct "where" an application runs, especially as one considers things like geographically distributed vMotion, or VDI for desktops, as examples.
This creates a unique opportunity to seperate the contents of the application container (what users really care about) from the details about how and where it runs (perhaps a cloud?).
The bet here is that most business users will care far more about the former, and far less about the latter.
A virtualized customer (e.g. potentially anyone running on any vintage of an Intel instruction set) now has choice -- run it on my desktop, run it in my datacenter, or ship it off to a vCloud provider to run it on my behalf. Or any combination of the above -- and be able to change my mind if circumstances change.
The user experience will be largely the same if the technology is done right. So it gets down to who can do the best job of running my container at the lowest cost, doesn't it?
But What About Economies Of Scale?
There's no arguing that a multi-tenancy environment can be architected from the ground up to provide astounding levels of resource and operational efficiency. And I'm not saying that any proposed vCloud would offer the same absolute degree of peak efficiency.
But we shouldn't underestimate the substantial potential efficiencies of VMware at scale. CPU efficiencies through hypervisors. Memory efficiencies through page sharing. Dynamic load balancing and service delivery management. Network saving through I/O consolidation. Consolidated operational models that are largely "hands off".
Going farther, one way to back it up. One way to provide business continuity. One way to secure it.
And, just maybe, some interesting storage efficiencies as well :-)
The bigger the environment gets, the more potential exists for efficiencies at scale.
And I think it'll be hard for many IT organizations to get to this sort of efficiency model -- simply because they won't have the scale that dedicated vCloud providers will enjoy.
What About Things Like Security and Management and Service Levels and ... ?
When looked at in the cold light of what's possible today, there are very good answers to all of this. Using today's technology, the entire virtual machine -- and all its information -- could easily be encrypted end-to-end. Model-based management technology can provide IT operations with a "portal" to provide service level management of the subset of resources that they control at the vCloud provider.
I am of the opinion that this has happened before.
You'll hear larger enterprises talk about their "corporate network". Well, as we all know, they don't own the cables in the ground, nor the switches in central offices. They work with AT&T or BT or whoever to build a logical corporate network where they can manage as much -- or as little -- as they want to.
They're free to tailor service levels, opt for advanced security, build networks from multiple providers, or combine stuff they own with stuff they rent -- if they want. It's all, well, virtual, isn't it?
Now, if we extend networking concepts to compute and storage, is it really all that much different? Especially when the result is essentially transparent to the end user if done right?
Makes the whole "enterprise cloud" thing seem a bit less futuristic, and far more pragmatic, IMHO.
Imagine This ...
One day, an IT service provider comes calling on a small-to-medium sized business. The specialist runs a P2V utility on all the production applications running on various servers. Another utility ships off the data to a remote location, maybe via a backup.
A couple of network reconfigurations, and -- presto! -- everything is running exactly as it was before. Except that users can't tell where the processing is getting done -- desktop, data closet, cloud -- and it doesn't really matter, does it?
Who Will Be The Players In These New vClouds?
VMware, of course, as a technology provider at both the customer side and the service provider side. They seem to be doing a pretty good job of getting everyone to put application stuff in virtual containers.
Service providers and outsourcers, of course. It's hard to tell what's real from the press release, but it's a pretty extensive list, and I bet they're coming to some of the same conclusions expressed here. I'll be waiting for some of these branded vCloud services to hit the market.
Infrastructure vendors, of course: servers, networks, storage, backup, etc. No surprise, EMC sees itself playing in these environments -- especially at scale.
But I think that two topics -- management and security -- will need to bridge the domain between how the service provider runs their vCloud farm, and how their clients can control certain aspects of the end-to-end environment, much like is done in corporate networks today.
I'd offer the opinion that -- architecturally speaking -- both Smarts and RSA are reasonably well positioned to play this role going forward.
And Who Will Get Disrupted First?
I keep coming to the conclusion that vendors who sell hardware to SMBs and the like will probably see the first impact of this "enterprise vCloud" trend if it happens in full force.
Based on my impressions, all these customers really care about is running their businesses -- period.
And if some nice server provider came to them and said "you get to do exactly what you're doing today, but you don't have to own IT hardware anymore" -- well, I think that proposition will get some serious attention.
What do you think?
Courteous comments welcome as always ...
Comments