Clouds Need To Be Better Than The Environments They Replace

You say something often enough, it becomes a personal meme.  If enough people agree with you, and start saying the same thing, it might become an industry meme.

In that spirit, let me share with you a meme that shows every sign of making that transition.

Why?  Because I think people are ready to accept this particular thought as conventional wisdom.

An Oversimplified History Of Cloud Thinking

The first round of people talking about cloud were the technologists. 

The message? "Clouds Are Cool".

These people were attracted to the power of the underlying technology to provide highly elastic infrastructure (as well as associated applications) that were largely free of traditional physical computing constraints.

Nice geeky conversations resulted, but -- if you weren't that sort of technologist -- all of this kind of left you somewhat confused and disengaged.  Plenty of those discussions still going on, though ...

The next round of people talking about cloud were the business people. 

The message? "Clouds Are Cheap". 

Hundreds (if not thousands) of studies were done, each comparing the cost-to-serve for a cloud-based service as compared to a traditional IT approach.

Many people found out just the opposite.  If you're doing enterprise IT at any decent scale, the convenient by-the-drink pricing models associated with so many external services can't be justified as significantly less expensive.

And that makes a certain sense: at decent scale, we all pay pretty much the same prices for IT inputs: hardware, software, power, labor, etc.  The difference?  External service providers need to make a decent profit ; enterprise IT organizations don't.

I think we've now entered the third (and most important) phase of the conversation: clouds (especially private clouds) can be a fundamentally improved way of providing enterprise IT services. 

The message? "Clouds Are Better" -- or certainly need to be.

A Bit Of Context

Given all the cloud chatter out there, I feel obligated to define my terms once again.

By "cloud", I mean any next-generation IT environment that's (1) built differently -- dynamic pools of virtualized resources, (2) operated differently -- purpose-built zero-touch or low-touch operational models, and (3) consumed conveniently -- use of resource includes pay-for-use models, or other forms of convenient consumption. 

Do all three, and you're a cloud in my book.  It has nothing to do with physical location or ownership model.  By way of example, a truck is a truck regardless of where it's parked, or who owns it.  BTW, you're free to disagree with this oversimplified view.

By "private", I mean "an environment under the control of enterprise IT", whether those resources sit in a traditional data center, or use compatible service provider resources, or -- ideally -- a combination of both.

Put up a cloud like environment that's completely under the control of IT, (regardless of where it might physically sit, or what name is on the side of the building), and I'll call it a "private cloud".

From "Cool" To "Cheap" To "Better"

I will argue passionately that -- in the history of enterprise IT adoption -- newer models have to be better than the ones they replace or augment.  It's not enough to be "cool".  It's not enough to be "cheap".  It has to be "better" than what it purports to replace.

So much of the cloud discussion tends to overstate the economic benefits, and understate the inherent compromises.  The thinking tends to be "cheap and cheerful computing for stuff that doesn't really matter to the business."

I beg to differ.

I believe that near-term cloud models (especially private cloud models) can be far better than the traditional physical environments that the  hope to replace.  Yes, they'll be more cost-effective -- but that's mere table stakes.

They'll be able to deliver a far better class of enterprise IT services than their traditional physical counterparts.

"Better" Is In The Eye Of The Beholder

When considering enterprise IT services, the definition of "better" tends to vary depending on who's doing the asking: the enterprise IT organization, or the people using their services.

That's unfortunate (aren't they really the same thing?) but it's the world we live in ...

The user of the service wants speed and flexibility -- at a competitive cost.  They're more interested in what the enterprise IT services can help them do for the business, and care less about the gory details.

The traditional enterprise IT organizations usually starts with cost, and then dives into all the gory details around implementation, compatibility, security, compliance, etc.

Both ends of the "better" spectrum are incredibly important.

So, let's look at just a small sampling of the ways where private cloud architectures can be potentially better than the traditional physical approaches they are attempting to replace.

Chuck's Quick List Of "Private Clouds Can Be Better Because ..."

Let's start with cost-to-serve.  Optimized infrastructure for virtualization that's designed for purpose can serve more virtual machines than infrastructure that's not designed for purpose.  That's one of the attractions of the Vblock and its associated technology components.

The right way to measure cost efficiency isn't to disassemble the components into its constituent parts, and then argue about who's got the best/cheapest server, network, storage, management interface, etc.  Instead, it's about measuring the "all in" cost per VM delivered.

Let's move on to time-to-serve.  Newer operational models can change the reaction time for providing a new service level (whether this is provisioning a new instance, or changing the service level for an existing delivered service) to minutes or seconds -- usually without IT intervention.  Not only does the business get what it wants far faster, but they consume exactly what they need at any given point of time, and do so without the need for expensive and cumbersome internal IT processes.

Now, let's talk about availability and recoverability.  The range of different potential service levels around availability and recoverability available in a fully virtualized private cloud model far exceeds what's available in the physical world -- and tends to be better, simpler and less expensive.

More importantly, the availability and recoverability attributes associated with a given set of applications are dynamic in nature -- they get applied when needed, rather than uniformly.  Try doing that with a traditional physical IT infrastructure approach!

How about security and compliance?  The whole story isn't out there (yet), but I will go on record that fully virtualized environments can be made far more secure, far more compliant, far more auditable -- hence far more "trusted" than *any* typical physical infrastructure approach.

The virtual machine gives us a convenient "virtual perimeter" to defend that dramatically simplifies so much of what is daunting in the security and compliance world.  Watch this space closely, if you're a security architect.

Software and application compatibility?  Basically, if it runs on Intel, it can run on a private cloud without any modifications.  The stack above the CPU is completely up to you -- operating system, database, middleware, application, etc.  Again, better than what we typically get in the physical infrastructure world.

The list goes on: resource management, patch management, licensing compliance -- pick your favorite pain-in-the-butt IT discipline, and I can make a case that it's far better in the fully virtualized private cloud world.

But there's more -- there's consumption models.  Embrace a private cloud approach, and you'll have a new set of options to either run your own infrastructure as efficiently as a cloud provider, use any number of a growing set of compatible service providers, or any dynamic combination you might need at a given point in time. 

Getting this flexibility in consumption model is something that's notoriously difficult in the traditional physical infrastructure world -- and doing so dynamically is near-impossible.

And The Business User Perspective?

It's simple: IT infrastructure doesn't get in the way of what the business wants to get done.  It's available when needed, conveniently priced, and can be changed at a moment's notice.  Sort of like power, or phone, or transportation, or any other form of infrastructure that business use.

Why should IT infrastructure not evolve to be similar to other forms of infrastructure?  No one asks me to pre-provision a phone call to Japan six weeks before the call, for example :-)

The answer is simple: that's exactly what is happening before our eyes.

Getting To Better

So much of the cloud discussion seems to embed an apology for one kind of compromise or another.  Hey, we're at three 9's availability.  We don't lose your data that often.  Our unique software stack isn't so bad.  We do an OK job at security and compliance.

I find all of this completely misdirected.

We -- as an industry -- should have as a goal to make our enterprise cloud environments -- whether they be private, public, etc. -- far better than the environments we are replacing.

Private clouds aren't all about cheaper IT, they're about better IT.  And being cost-effective is just the start of the discussion.

Comments

[nate]

I agree, with the right scale you can often get some pretty cool things to scale out, I came across this cloud provider on Marc's blog a couple weeks ago and was incredibly impressed with the thought they put into their infrastructure, and they are quite proud of it of course by giving a lot of details on how things work internally. I haven't talked to them myself just read the details on their site -

http://ultraspeed.com/managed-hosting/
http://www.ultraspeed.com/technology/rack/
http://www.ultraspeed.com/technology/server/
http://www.ultraspeed.com/technology/storage/

The only thing they don't elaborate too much on is their network, maybe they are still working on making something worth bragging about :)

I've been playing with Terremark vCloud express for my own private needs, certainly not cheap, but I think affordable enough given my meager requirements(1 CPU, 1.5GB ram ~50GB HD). I wouldn't use them for anything at scale for cost reasons. But they have good infrastructure too(EMC partner too). Kind of ironic they list Intel as a partner yet as far as I can tell all of their servers use AMD CPUs(from an employee response to a question asked in the forum and my own VM is running on an Opteron 8389)!

Maybe Intel paid them off to make their next generation of gear Intel-based.

[Chuck Hollis]

Hi Nate -- all good!

I'd encourage you to share your thoughts here, or (ideally) on a separate property we've put up for these discussions: www.privatecloud.com

Either way, we want to hear about your thoughts!

-- Chuck

[nate]

I will keep my eye on it! It certainly looks very formal(lots of suits and stuff), not sure how well I might fit in, but if I see something you know it's hard to self restrain from commenting :)

keep up the good work..