“I’ve looked at cloud computing from both sides now,

From up and down, and still somehow,

It’s cloud illusions I recall,

I really don’t know cloud computing, at all.”

-        (With apologies to) Joni Mitchell, 1969

 I searched for the word “cloud” on Google recently and it returned 762m hits. It wasn’t until I got somewhere in the second page of searches that something relating to weather came up. Weather aside, what, really, is “the cloud”− much less “cloud computing”?

To answer this seemingly simple question I referred to Wikipedia, which defines “cloud computing” as, “… the delivery of computing as a service rather than a product, whereby shared resources, software, and information are provided to computers and other devices as a utility (like the electricity grid) over a network (typically the Internet).”

OK, I get that. But why is that different from timesharing computing, hosting, or grid computing? More importantly, if I’m a business owner or IT manager looking to implement a cloud solution, why do I need to know the answer to that question?

Let’s take a look.

 

How We Got Here

Timesharing computing was similar to cloud computing from a usage point of view but implied access to one computing device that was shared, not a range of computers that were connected together. You tended to “dial-into” to use one application and signed off when you were done.  However you connected to the resources, they were not re-deployable to other users and your data was not portable.

In hosting, you still used a service that either “hosted” or ran your application, or perhaps you used a generic version of an application. The ultimate version of this in my view is SalesForce.com. I am sure they would see themselves as a “cloud provider” − but isn’t it just an extreme sort of hosting?

Grids, on the other hand, tend to not have applications in them – unless you put them there. For years, the problems with distributed computing were the expense of the computers, the speed of the network and different computers running different versions of software. Over time, the industry has solved the expense problem and is a long way toward addressing network bandwidth challenges. And today we use open standards to address the problem of different computers running different versions of software.

While you don’t necessarily need to know the type of computer in “the grid”, you do need to understand the nature of the standards to submit a job.

Sounds technical – yes it does, and grids tend to be the domain of pure math, physics and engineering. They are designed to run specific kinds of applications that don’t share well – these applications tend to want all of the resources of the grid when they are running because they are solving big, hairy problems whose speed is directly dependent on the underlying compute and interconnect speed of the grid. And that means that most grids are designed to solve one class of problem and to run only one “job” at a time.

So that gets us back to “cloud computing”. Using the Wikipedia definition, all the other types of computing could be thought of as “clouds” too. My guess is they probably are.

What’s more, I don’t think it matters.

 

Cloudy Ahead

Let’s go further and argue that that definition is perhaps wrong or just insufficient. If we are going to understand the next generation of computing, while we can call it “cloud computing,” in reality we need to look less at the hardware and more at the actual workload – or, more to the point, to the methodology that is being executed with cloud deployments.

When we look at how the methodology is being executed, we see some things that are common and consistent: easily provisioned systems, virtualization, power efficient dense environments, flexible platforms, a focus on agility and dynamic allocation.  When you consider these factors in varying combinations, you see the broad range of solutions that can be called “cloud”:

• In today’s Web 2.0, social-networked world, we readily consider Facebook, LinkedIn, Twitter and Amazon as “cloud” destinations. Each builds on a classic three-tier architecture, creating a large system footprint − but with industry-standard x86 technology. 
• With Web-based business applications offered by companies like Salesforce.com and Google, the need to run applications locally is removed, allowing data instead to be stored and manipulated “in the cloud”. One impact of this is that updates can be deployed instantaneously, driving change in client computing.
• And many companies today are turning to the “private cloud” to allow them to be more agile, respond more quickly to business changes and improve return on investment. In some cases, business units are being empowered to self-provision and deploy servers. Think of this as virtualization on steroids: where virtualization moved the deployment time of servers from weeks to hours; with private cloud it moves from hours to minutes.

 

 My Cloud, Your Cloud

How you look at the cloud is, of course, also a function of whether you are a user or a provider.

As a cloud user, I shouldn’t really care how the content is provided − I care that it’s scalable, flexible (no big upfront definitions of what it can and cannot do) and is “on demand”.  Mostly I just want it to be variable – in how I use it and pay for it, and in the applications I push to it.

But if you are a cloud provider, you want to give your customers the service levels they want − and the variability they demand. If your cloud model is going to be priced right, you will need to smooth demand over time, be both ultra-available yet be able to accurately charge for real-time usage. The real cloud providers don’t do single-user clouds, they do multi-use or multi-tenant clouds − and they design their clouds to address a specific range of applications very well.

The trick for these cloud providers is to minimize costs while ensuring customers’ applications run as promised; completely generic hardware infrastructure won’t meet your goals.

 

One Size Does Not Fit All

These cloud providers will not be like Google and Amazon − and that’s not a surprise, right?

Many, if not most, will be about small applications; in a previous world we would have called these “Objects”. These are packages of application and data that travel together and can be executed around a virtualized infrastructure. When we explore these packages we see some common elements:

• They are based on open standards, which can be executed on a wide range of computing devices and in a wide range of locations.
• Sometimes these work packages will be done remotely, sometimes locally and more often than not at both places.
• The data will need to be accessible regardless of the device or location from which you want to access it.
• The applications will therefore need to be highly-threaded and work with multiple types of processing cores (given that one size tends not to fit all).

While we like to believe that everything works the same on open standards, the truth is that too many options makes life too complex. Therefore we can start to make some assumptions about the basics of what will run the cloud.

From a hardware perspective, there will be various types of cores but I’m willing to bet that x86 will be the dominant processing platform for servers and mainstream clients for many years to come – and yes, ARM for ultra-low powered clients. While there may be other type of servers (including maybe one day ARM-based ones), and there may be other types of low-powered clients, in reality computing moves much more slowly than the media might have you believe. Just check the size of IBM’s mainframe business today – smaller but still very much there.

In terms of operating systems, you would have to bet on four main players – Windows, Linux (or the LAMP stack), iOS and Android. What about WebOS and RIM? Like IBM’s mainframes, there may be a place for them but it will not be the main focus.

In terms of applications programming and languages – how things have changed. Three years ago we would have said Java, Ruby and some others would make the cloud go. While they may play a huge part, I suspect HTML5 is the real one to watch. I think we are just at the start of this revolution.

And I could go on and talk about databases and other middleware, but like all the other parts of the stack, they will consolidate too.

 

Ease of Engagement

So where does this leave AMD? Right where it has always been – which now seems to be in the middle, edge and the ends of the cloud. I don’t know who the next Google or Amazon will be, but – if the current trend holds – it will rely on AMD.

Chances are you got to this page through “the cloud”, and if so you used AMD along the way. You may be reading this blog on an AMD VISION Technology PC, or through our AMD Radeon™ graphics technology and it may even be hosted on a server powered by the new AMD Opteron™ processor.  

AMD is at the heart of your favorite social media sites, search engines, application providers, network providers, device providers and even a large chuck of gaming consoles.

And AMD technology is designed to provide you with a better overall experience, from making those devices run faster, to enabling longer battery life, to creating superior visual experiences. From the AMD Fusion Family of Processors in your PC to the AMD Opteron-based multi-threaded servers, “the cloud” may feel like an illusion − but it’s made real with help from AMD.

 

 

Nigel Dessau is Senior Vice President & Chief Marketing Officer for AMD. His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for the contents of such linked sites and no endorsement is implied.