Press and Analysts React To Private Cloud

I've worked with the press and analyst community for many years. 

Say what you will, but I've found them a very good barometer of what ideas will fly, and which ones won't. 

Sure, they don't actually buy IT products.  But they do get exposed to just about every IT pitch in the world, and they usually have a good nose for what's going to be a popular discussion, and what won't.

I've now had a few outings where I've had an opportunity to share private cloud concepts with these folks -- and how VMware, Cisco and EMC are working together to accelerate this vision.

The reactions have been mostly favorable, but with a few interesting wrinkles here and there.

What Is This Private Cloud?

You may have seen me write about this many times -- for a quick overview, please see here -- and if you're interested in more, you can see all my posts on this subject here.

If you'd like to see some press and analyst quick takes resulting from last week's EMC World, check out this, this, this, this, this and this.

In a nutshell, a private cloud describes a fully-virtualized enterprise IT environment.  Applications, desktops, supporting software, etc. are all neatly wrapped in virtual containers. 

Operational style is biased towards on-demand and self-service, and less towards static provisioning and management.

And, since everything is nicely wrapped in a virtualized "shipping container", private clouds also allow compatible service providers to offer IT infrastructure resources as an alternative to owning the stuff yourself.

It's a powerful, simple and comparatively pragmatic view of enterprise IT -- once you get your head wrapped around it, that is!

That Isn't A Cloud!

The first -- and most frequent reaction -- from people is simple intellectual rejection: what you are describing is not a cloud.  I actually welcome that reaction, since I'm prepared to go a few rounds back and forth.

One hangup some people have is that if parts of it run inside a data center, it can't be a cloud.  I beg to differ.

For example: take a pile of servers and storage, wrap it in virtualization, manage it in an on-demand, dynamic and self-service fashion.  Most people would agree "yep, that sort of sounds like a cloud". 

Would it really matter that much as to whose data center it runs in?

Another frequent reaction is "this doesn't sound like Amazon and Google and whatever that Microsoft thing is ...".  Yes, they're right.  I would say that those are examples of "public clouds" -- open to all comers -- rather than "private clouds" -- under the explicit control of IT departments.

OK, So Maybe A Private Cloud Really Is A Cloud

By now, I've usually got them agreeing that being a cloud has very little to do with geographic location or the underlying business model of the service provider.  Neither of those do a cloud make.

The discussion inevitable turns to comparisons with other enterprise cloud models that are out there.

One is application compatibility.  Private clouds -- since they use x64 virtualization -- can potentially 'cloudify' any legacy application that can run on an Intel (or AMD) processor.  No need to rewrite or convert.  That largely isn't true for many of the other public cloud models.

Another is a very wide range of infrastructure and provider choices.  In a private cloud -- since everything runs in a standard virtualized container -- enterprise IT organizations are free to run these virtual machines on their own infrastructure or using any number of third party providers that support compatible technology.  This largely isn't true for many of the other public cloud models today.

And finally, the private cloud model assumes that enterprise IT is still in control -- service delivery, security, resource usage, etc.  The only thing that changes is new options as to where the infrastructure might live -- your data center, or someone else's.  Again, not largely true for many of the other public cloud models today.

What About IaaS, PaaS, SaaS ... ???

And then there's all these "*aaS" variations as part of the discussion.  How do these fit in?

I believe they all can be construed as potential ariations on a broader theme of private clouds, as follows:

Infrastructure as a Service (IaaS) -- think compute, storage, networking, etc.  I tend to think of Amazon's S3 and EC2 services in this category.  All of these services will provision a "virtual container" where you can put your stuff. 

Typically, these environments presume you're building a new application.

I would argue that a "virtual machine on demand" service -- with your choice of tools and operating system -- qualifies as infrastructure as a service.  At least, how it's commonly defined today. 

Now, put that virtual machine under direct control of IT -- or even stand up your own "infrastructure as a service" environment behind the firewall -- or do both.  Either way, you've built a flavor of a private cloud.

Platform as a Service (PaaS) -- think about outsourcing in a virtualized world.  Enterprise IT already has customized applications or desktops running comfortably in virtual machines -- and just needs a better/faster/cheaper place to run these containers.

Enterprise IT now has a choice -- run them on "owned" infrastructure behind the firewall, or run them on a compatible service provider outside the firewall (still controlled, managed and secured by IT), or any dynamic combination of the two.  Once again, a private cloud.

Software as a Service (SaaS) -- this one is extremely interesting to me.  Consider Salesforce.com for a moment.  With their current model, there's only one way to consume their software -- as a service.  If you want to run their software in your data center, you're flat out of luck -- aren't you?

Now, imagine a model where that software ran in virtual containers.  You'd have some interesting options, wouldn't you? 

You'd be able to choose -- dynamically -- whether you wanted to run on your own infrastructure, or someone else's.  As a matter of fact, you'd be able to choose between many external service providers, rather than only one.

Indeed, just about any software title can potentially be "SaaS-ified" using a private cloud model.  The value moves to the application and its information, and less about exactly where the infrastructure might live.

OK, I'm Starting To Get It, But ...

By this point in the discussion, most members of the press and analyst community now understand the concept of a private cloud, and -- more importantly -- how it compares and contrasts with other cloud models being offered to enterprise IT organizations.

And then the next round of questions and objections start to come out

For example: don't these public cloud providers offer economies of scale? 

The answer is potentially "yes", but with an important caveat -- these are purpose-built environments that do a few things very, very well.  Like "search".  And "email".  They are not suited to the sort of general-purpose IT most enterprises need, and would have a very different cost structure if they were.

And they're most certainly not designed to run traditional IT applications, and federate with both enterprises and/or other service providers.

Or: what about security?

Many of us have become convinced that fully virtualized environments can be made far more secure than their physical counterparts. 

Besides, if you think about it, we've already learned to get comfortable with other shared infrastructure environments: outsourcing out of common data centers, sharing network trunks, etc.

So What's Holding Things Back?

This is the great part -- not only are the key technologies already in the marketplace, but enterprise IT shops are already doing the pre-work required to build private clouds: they're virtualizing their applications.

Sure, we could see better and more mature technology solutions, and -- of course -- there's lots left to virtualize in the enterprise IT landscape, but already we're seeing large IT shops who've virtualized hundreds or thousands of application images -- and are thus poised to take the next step forward.

The question of standards often comes up.  That's a tough one.

We'd all like to live in a world where we have all the de-jure standards we need to feel comfortable that every component from every vendor is going to interoperate perfectly all the time, but that's going to take a very long time indeed.

In the meantime, a few key standards are in place (OVF -- the open virtualization format -- comes to mind), but we'll need to see a real commitment from the vendor community to make sure things are very interoperable indeed.

And Where Does That Leave Us?

The part I like is that I believe that private cloud concepts are now an essential part of two very important discussions.

One discussion is around how various cloud models will serve the needs of enterprise IT.  The private cloud model stands in stark contrast to other cloud models already part of the discussion, and that's a good thing.

The other discussion where private clouds play is the ultimate question on the minds of most IT thinkers: "where is enterprise IT going?". 

And, for the first time, I think I really know the answer to that question.

Comments

[William Vambenepe]

When you have a hammer everything looks like a nail. When you have a hypervisor everything looks like a VM.

[Chuck Hollis]

William

Ah, yes -- so true -- it's human nature.

It's also human nature (especially in IT people) to be inherently skeptical of anything new that challenges their conventional thinking.

We'll see what happens here, won't we?

Thanks!

-- Chuck

[Prashanth]

seriously, it seems like marketing gibberish. i've been following your blog for a long time and this beats everything else to the top for some gibberish :-). and btw, there are other posts which i really like...

as an IT professional, i still cant make sense of it - it would help if you pinpoint a concrete example.

[Simon Munro]

I used to be part of the ‘private cloud is not a cloud’ camp because if felt that the removal of ‘public’ from the cloud also removed the abstractions necessary to create solutions that are truly cloud capable – location, network and security agnostic and where some real thought needs to be put in to the architecture.

However, I realised that a big part of the cloud is the administration and rapid provisioning of resources rather than a SOA utopia or where growth is expected to be that of Twitter. I think that the adoption of cloud oriented technologies within the enterprise data centre is going to be a major driver of adoption of the cloud – the lag on public cloud application architectures (and related skills) will mean that the initial cloud focus will be private cloud anyway.

Simon

http://blogs.conchango.com/simonmunro/archive/2009/04/17/the-cloud-is-a-response-to-demand.aspx
http://blogs.conchango.com/simonmunro/archive/2009/03/10/bring-on-the-private-cloud.aspx

[shiningarts]

Metaphorically, we all are becoming technological meteorologists of sorts these days: Since Sun set into Oracle, the ubiquitous Clouds have moved in. Just as we all want to enjoy private waterfronts, we want to privatize the Clouds so that we can control certain private information and data discreetly. Fortunately, there are plenty Clouds out there for everyone to enjoy. From my vantage point, I believe that the next movers and shakers are not going to come out of database, server, OS, programming language, networking, or storage business. The next major development on the horizon will be led by someone who can put these together efficiently into a Cloud, so that the atmospheric water vapors condense into drops heavy enough to become rainmakers. Definitely, it will be an evolutionary process. No specific software or hardware will be retired because they are considered as old legacy systems, old-fashioned, or outdated. They will do their share of the Cloud Computing directly or indirectly as long as the touch points/connectors are Cloud compliant. Because of the shortcomings of the existing bandwidth and implementation of the current Web 2.0 technology, I think the defacto and practical implementation of Cloud Computing will be through the virtualization for now. It may change as new technology emerges and becomes available in the future though. Virtualizing clients in addition to servers could be a good approach to overcoming the underlying bandwidth issue and to promoting telecommunications--wouldn’t that make those telecommuters who like to promote the ever-expanding green initiatives happy!. All in all, it's all about how to horizontalize or virtualize the disparate vertical silos that are prevalent in our current environments. As long as we don’t experience any unforeseen atmospheric turbulences such as tornados or hurricanes, I believe, Private Cloud is on the right track.