Andi Mann at CIO.com posted a great article on challenges in adoption of virtualization and proposed a new term - VM stall. It refers to the apparent limit in adoption of virtualization within companies — those who start tend to virtualize the low hanging fruit and virtualization efforts seem to stall around 20-30% of the applications. Read his article

I agree with Andi that virtualization is not as widely adopted as everybody makes it look. By now, most IT shops have done something, so there is a widespread notion of “everybody’s doing it”. Also, it tends to be picked for new projects (which is a good thing).

I don’t think we should set “virtualization goals” — say at 50%, 90% or 100%  - virtualization is not an end in itself. That said, it is an essential enabling technolgogy and I do believe it will become the norm within the next few years — every new server will be virtualized from the start; likely it will be either shipped this way by the hardware manufacturer or the OS will provide the layer by default.

That said, the let’s look at the possible reasons for the current stall in adoption. Based on my experience, I would propose three:

1. Not enough value in virtualization alone: Server virtualization as a solution is mostly about server consolidation. As a result, companies tend to consolidate the non-critical apps, the ones that can be packed in small boxes. Once that is done, the value of moving the bigger apps is simply not there — there will be no consolidation benefit. The bigger apps need more resources, not less. Virtualization alone does not help much (and can make things harder in some cases if not properly implemented).

2. Cost: To get most of the virtualization benefits, you end up needing very expensive hardware — SAN for everything, fast storage interconnect, lots of network bandwidth, etc., and lots of software licenses, both for the virtualization and for the management of all the pieces. Many projects simply can’t justify that cost. Not all virtualization products have this characteristic, but the default choice does, so it skews the statistics.

3. Complexity: the mere mention of a “virtualization team” (in IT departments) shows that the technology, in its current incarnation, is not ubiquitous enough. Virtualization was not (and is not) supposed to become yet another silo. All IT professionals should be skilled in virtualization. If the technology is so complex that it requires separate virtualization team, then we need better technology.

Best regards,
- Peter

PS The solution to the above is in cloud technologies, where virtualization is an enabler to further abstraction and encapsulation. Good cloud technology - like the one we have here at CA|3Tera — achieves simplification and flexibility that we have not had before. This helps overcome the 3 factors I listed above and move adoption beyond the stall limit. Watch this space for further posts on how this happens.

I don’t know why, but I am still surprised when I hear the following question. What’s the difference between virtualization and Cloud? To me, it’s like asking the question – What’s the difference between a hammer and carpentry? The latter is a comprehensive craft. The former is one of many tools used by the craftsmen who practice it.

Simple – right? So why does that question occur at all?

It occurs, in my opinion, for two reasons, one right and one not so right.

The first reason is that all of the server virtualization vendors of any significance are also introducing Cloud offerings to the market. So, people are naturally associating the two (and rightfully so, just like one would associate hammers and carpentry). The difference is, though, no one thinks hammers and carpentry are the same thing.

So, the not so right reason – There are Cloud computing laggards out there who would like us to think that virtualization and Cloud are similar because they have embraced virtualization technology and do not want to appear out of step. As a result, there is a ton of noise in the market that is very hard to sort through.

So, how do I suggest one sorts through this noise?

When faced with a potential Cloud solution, ask a few questions about it.

Does it help me provision and deploy virtual machines on demand? If the answer is no, I’d ask why are you even looking at it? But if the answer is yes, just deploying VMs on demand does not a Cloud make.

Does it enable the encapsulation and on demand deployment of multiple VMs as a single entity? If rather than managing VMs, you want to manage frequently used “appliances” that are comprised of multiple VMs (e.g. a specific app server, a specific messaging system and a specific database server), can you do it? If the answer is yes, you are on your way to a real Cloud solution.

Does it enable the encapsulation and on demand deployment of whole software stacks (e.g. LAMP, Ruby on Rails, .NET, etc.)? If the answer is yes, you are certainly in the Cloud.

But, do you want more? Does it enable encapsulation and on demand deployment of entire multi-tiered apps? If yes, you have a very powerful Cloud solution.

More? Does it enable the encapsulation of the apps along with everything they need to run – network, storage, infrastructure, configurations, policies, documentation, etc., etc., etc.? If yes, then you have the most complete Cloud solution of all.

So, you might sense a theme here – Encapsulation. Yes. Encapsulation is key, but it is only half of the story. Encapsulation itself results in many benefits, especially operational cost savings and decreased time to market. But encapsulation alone does not make a Cloud. It does not create portability. It does not create the ability, by itself, to deploy anywhere, any time.

What’s the second half of the story? Abstraction. Not only do the most comprehensive Cloud solutions have to provide unlimited granularity of encapsulation, but they must completely abstract what is encapsulated from the physical resources (machines) they run on, so that they can run anytime, anywhere there are available idle resources.

In short, you do not measure a Cloud solution by how it does virtualization. You measure it by the granularity of its encapsulation capabilities and its ability to abstract VMs, stacks, apps/services and entire data centers from the physical resources they run on.

So, what is the future of virtualization and where is it going in 2010?

Virtualization is going the way of the hammer. It will be a necessary commodity for the Cloud, just like the hammer is a necessary commodity for the carpenter.

Now, before all the virtualization vendors get their shorts in a knot and start screaming at me that I am implying that all virtualization is the same, I am not. I acknowledge that some have features others do not, some outperform others, etc. But, can you tell who the best carpenter is only by knowing what brand of hammer he uses?

Ken Fogarty, writing for CIO, comments on a panel on virtualization at MIT last week that included Amazon CTO Werner Vogels and VMware founder Mendel Rosenblum.

“The good news is that virtualization will become a critical part of an even larger part of most IT infrastructures as time goes on.

“The bad news is that it will do so as part of a larger movement toward cloud computing and will, in large part, disappear as a separate discipline.”

This trend started a year ago and IMHO is new happening faster than most folks expected. Perhaps this explains why so many vendors seem to need to claim they are in the cloud computing space.

This week, something quite miraculous happened.  Those of us whose vision of the future is in the Clouds have seen our crystal balls start working.

Months ago, 3tera unveiled our Cloudware architecture.  But rather than try to convince the world that there is only a single architecture that works and ours is it, we emphasized that Cloud architectures need to be open.  Not only need they interoperate with all sorts of hardware and software as virtual appliances, they need to interoperate with other Clouds and Cloud components as well.

So, what happened this week?

The two undisputed leaders in virtualization, VMWare and Citrix/Xen announced suites of products in support of Cloud Computing, vCloud and Citrix Cloud Center (C3), respectively.  Undoubtedly, Microsoft and Red Hat and more will follow.

The anticipation that drove our Cloudware architecture is proving spot on.  There will be multiple global Clouds, they will not all be the same, and the ones that will get the brass rings will be the ones that interoperate rather than stand alone.  Cloudware is designed so that it will, in the not too distant future, have the ability to incorporate elements from any Cloud.

You will note that both VMWare and Citrix, in their Cloud announcements, emphasize the need for API-based interoperability among Clouds.  3tera agrees.  The development of this interoperability will make the vision easy to accomplish.

3tera intends to take this direction to the nth degree by not only enabling applications in one Cloud to interoperate with applications in others, but to enable elements from multiple Clouds to coexist in the same application.

There’s been a lot of music to our ears this week.  The huge install bases of VMWare and Citrix are becoming part of the eco-system that we have been participating in the definition of, designed for, been building for and support - the eco-system we’ve been predicting was coming.

So, if this is music to 3tera’s ears, it’s a multi-media extravaganza that should tickle all the senses of information technology users of any size   Combining a continued direction of open Cloud Computing where anything can operate in the Cloud with this new direction of interoperability among Clouds will leave all IT users at their own mercy.  Vendors will have less ability to manipulate and dictate what hardware you run your applications on, what operating systems you use, what software you deploy, what type of infrastructure components you rely on, what databases you use and where your applications run - and in how many places.  This will ALL be up to you and you’ll be able to change it all at YOUR will with just your little old browser.

Continuing on yesterdays theme of vrtualization supplanting operating systems,
I came accross an article today on Information Week titled
Hypervistors May Replace Operating Systems as King of the Data Center.

I’ve written for some time now that the role of the operating system is changing, as I noted in yesterday’s post. However, I have trouble with the author’s premise in this story. As I commented there:

“Hypervisors possess neither of the two factors make the OS strategic. First, software applications are not written directly to APIs specific to the hypervisor. Instead applications still run on Windows, Solaris, Linux or FreeBSD. Since all the development tools come with the OS, that seems unlikely to change. Second, users are beginning to understand that managing a data center full of VMs is a nightmare, prompting one blogger to refer to virtualization as the ghetto of the data center. This has created a market for VM management tools as the acquisition of Xensource yesterday points out.”