Is There A Cloud In Your Future?

I am at the stage where I'm constantly thinking about a new concept.  It's burning a hole in my head right now, it is.

In a few months, I'll probably be able to articulate it clearly and concisely, but right now I'm rambling a bit.

If you can bear with me during my learning curve, I'd like to start a discussion about what I'm starting to realize may be the *real* next-gen IT.

Something that makes things like SOA, grid, virtualization, etc. look like warm-ups for the main act.

Some Very Different IT Discussions Lately ...

I think it started about a year ago.  I was meeting an entirely different kind of customer.

They usually had what I'd call a "web 2.0" business model.  Lots of content.  Lots of users.  And a clever way of connecting it all up.

When I pitched EMC, I ended up missing the mark.  So I started listening, and their needs were very different.

Life At Internet Scale

These people had a business model that got really big and it worked well, or it didn't work at all.

Now, in the traditional IT world, we throw around the word "petabyte" to impress people.  Well, these folks were thinking in terms of a petabyte a month if everything worked out well.

From a storage perspective, it was a strangely different world.  Issues like backup, replication, availability -- these weren't big deals. 

OK, so you can't get to your favorite blog, as an example.

Costs were a *very* big deal.  Storage costs.  Admin costs.  Energy and data center costs.  Network costs. 

We're not talking about the "shave some money from the IT vendor" type of cost negotiations.  We were playing a whole new game -- where hitting the right cost structure could make or break their entire business model.

Why?  Take even a small cost, multiply it by a very big number, and you have a very big cost.

Networks Were A Big Deal As Well

Life at internet scale means that you don't put everything in one big honkin' data center.  No one can afford the network to do it that way.  Which means you need something like a virtual internet data center -- lots of points of presence, but it looks like one big honkin' thing.

And, for the real rich content versions of these customers, network delivery mattered as well.  If you're running a shoot-em-up gaming site, network latency REALLY matters.  Or if you're streaming HD video, as another example.

Having a predictable network service level was an essential part of their business model.

The Security Issues Were Entirely Different

They were concerned about being able to really know who they were dealing with at the endpoint of the network.  Being able to license and control content became important as well.  Some of them needed to be able to do real-time transactions, some of which were pretty high-value transactions as well.

It was a very different security discussion than I was used to.

And There Was More ...

... but I think you understand the gist of what I was seeing.  These folks were thinking about IT completely differently than I was used to.

I've gotten into the habit of calling this a "cloud" customer.  They're building a big, distributed, scalable and ultra-cheap cloud that others can tap into.

Well, that's just the new Web 2.0 crowd, right?

No, not really.

Because about six months ago, I started to meet more traditional businesses that started asking some of the same sorts of questions.  It took me a while to understand how their businesses could require something like this, and they used different words to describe the concept, but we ended up in the same "cloud" discussion.

They were looking at the same sort of thing -- not through a service provider, but as part of their mainline IT investment.  Whoa!

So, Now I'm Hooked

I'm starting to see clouds everywhere.  Not only in the latest-greatest flashy Web 2.0 companies, but in some pretty traditional names as well. 

I can't name the companies, but I can name the industries: health care, energy, manufacturing, retail, financial services -- have I forgotten anyone here?

On a personal level, I'd like to live as a user of these clouds.  Generally speaking, I don't want to own IT, I want to use it.  Just tell me where to point my browser, please.

I spend most of last year blogging about the core concepts of information infrastructure, at least how EMC saw it.

It looks like I'm going to have to revisit that discussion for "cloud" information infrastructure.

Because it's a very different discussion.

Comments

[jl]

This reminds me of my early EMC days being the lead technical guy for this huge ISP in northern VA whose name I will protect :-)

Anyway, in one datawarehouse environment they had 6 large Symmetrix arrays attached to - gulp - ONE big honkin host. This is not the model you're describing, but back then it turned the entire Symmetrix equation upside-down. The premise back in those days was 1 array, many hosts. This was 1 host, many arrays, and a scale that was eye-popping. I clearly remember the customer saying "the challange our vendors face is not the complexity of the environment, it's the scale."

Your blog clearly says "welcome to level 2".

[marc farley]

Nice post Chuck,

I've spent some time thinking about the integration of SOA with storage networking and you hit on a number of interesting points. I'd suggest that plasma is a better word than cloud because it infers a free and highly dynamic association of components within the system.

The applications would be web-enabled and developed using Javascript or something similar. The applets that comprise what used to be traditional brick and mortar applications could run on any virtual system located dynamically on any number of physical machines in the grid. Data would be accessed over a storage network that provides the means for any applet to access whatever data it needs. A back-end database or transaction system for all this is the lynchpin - and will probably take numerous iterations to work out the kinks.

Progress will occur in baby steps, because people won't want to take big leaps. In such an extended evolution there will be opportunities for new software companies to provide key parts of the puzzle. Crash consistent data protection for plasma computing would be a good example of a major software opportunity.

[Bert Armijo]

The early adopters of "clouds" built their own infrastructure from necessity. No vendor could help google, because because no vendor had never seen a customer with such needs. Yahoo, Amazon, eBay, YouTube, Skype, MySpace, and Facebook have all had to build it themselves. An effort that took years and hundreds of staff. As you might expect, each considers this a strategic investment and a barrier to entry.

New Web 2.0 companies don't have years, though. The experience of iLike shows they can reach internet scale in weeks or months. Even if they had the time and wanted to build it, the skillset required is so rare they probably wouldn't be able to find the people.

A new set of services has sprung up to service these companies. Amazon's EC2 is the most well known, exposing the system they used to build Amazon to other developers. At 3tera, we took a different approach. We leverage the hundreds of thousands of servers in the hosting industry, giving our users the ability to select and/or switch data centers whenever they choose. As more folks begin to recognize the space, I suspect we'll see more business models.

[Chuck Hollis]

I tend to see more of Bert's view than Marc's view.

The serious "cloud" folks really don't think about database the way the IT 1.0 crowd does. More simple indexing of metadata, rather than sophisticated data processing.

Bert's right about one point -- the variety and complexity of business models we'll see in this space will be staggering.