“Cloudy -- with a Chance of Meatballs”

This current kids’ favorite features a zany inventor whose offbeat creations either end up on the junk pile or terrorize the neighborhood. Until he develops a machine that can turn water into food, so weather takes on a totally different meaning and transforms his hometown’s economy, which was based on shipping canned sardines.

Transformation is also the theme of Cloud Computing.

Leveraging the underlying virtualization technologies which enable dramatic hardware consolidation, CapEX and OpEx reductions, and energy efficiency, Cloud offers new models for delivering IT services.

Arguably, Cloud computing will drive changes as extensive as IP did. Over-all, resources will be available just in time – instead of just in case. The promise is that over-provisioning in both networks and data centers will become a thing of the past.

That’s what’s driving the frenetic pace, as vendors, enterprises, service providers, consultants, analysts, investors all whirl around the cloud trying to define strategies, identify opportunities, and devise competitive advantage.

So what’s really different here? The issues and problems are the same ones the industry has been grappling with for decades: Application or service availability and performance, quality of end-user experience, rapid reliable creation and deployment of new applications and services, reduced OpEx and CapEx, security of access and of data….everyone’s familiar list.

When I was in college, mainframes were king and interactive computing was a gleam in a few people’s eye. Batch jobs were delivered to the data center in racks of punched cards, and results were available a few hours later. Gradually, timesharing came into vogue, with companies like GE Information Systems shouldering the big iron costs and enabling leased-line-connected customers to access computing power they otherwise could never have afforded. Then came off-site payroll and other systems, such as ADP offered. And then, of course, the flood of departmental and individual computing enabled by networking, mini-computers and PCs.

Aren’t we seeing a similar evolution?

The hulking mainframe in its air-conditioned, glass-walled sanctuary is now unlimited, invisible compute power somewhere “in the cloud”, where thousands and eventually millions of virtualized devices will provide on-demand and pay-as-you-use computing power, storage, applications, and services to billions of devices ranging from hand-helds to net-books and note-books, to wired homes and smart buildings, and even “smart cities”. The “end-user” may be literally anywhere, perhaps a person, a collaboration group, a community of interest – or in even more billions another device or application, such as an On Star automatic notification that you have just been in an accident because the airbag in your car deployed, also relaying location from GPS. Or a smart power meter that advises appliances in your home to reduce or stop their power consumption because over-all demand has exceeded thresholds and the regional grid has automatically “ordered” cutbacks to avoid brownouts and blackouts. Similarly, the service or application may be anywhere, and may change dynamically as conditions change. Face-recognition software monitoring international arrivals through CCTV at London Heathrow may be overwhelmed when using CCTV scans of Piccadilly Circus on New Year’s Eve unless expanded resources can be made available instantly.

A fascinating presentation by Google recently outlined its design goals: ~10⁶ - 10⁷ machines; ~10¹³ directories, ~10¹⁸ bytes of storage, spread at 100s to 1000s of locations around the world, ~10⁹ client machines and requires “automatic, dynamic world-wide placement of data and computation to minimize latency and/or cost" depending

on constraints such as bandwidth, power, resource use etc. See: http://www.cs.cornell.edu/projects/ladis2009/talks/dean-keynote-ladis2009.pdf

These are real differences of scale. But fundamentals remain: Deploying, monitoring, managing, securing -- and doing it all at affordable cost.

With such vastly increased scale and complexity, automation assumes new significance. Whether supporting the provisioning of a service chosen by a user at a self-service portal, dynamically moving applications among virtual machines based on demand and load, or provisioning additional capacity to meet flash crowds from successful marketing campaigns from, for example, an airline – automated processes are mandatory.

But to have a truly self-monitoring and self-adjusting cloud environment, which is required by the scale and scope of both services and users, automation must be based on as broad and deep collection and analysis of relevant information as possible.

Self-monitoring that feeds analysis and derives automated action to modify or recover. Again, fundamentals and evolutionary phases we have seen before.

So what’s the message?

Easily accessible shared or cloud-based services and resources, available flexibly, on-demand, in a self-service and pay-as-you-use basis is a very attractive model for providers and users alike. As a Gartner analyst said at their recent Symposium: Cloud is the top tech trend and issue for 2010.

Certainly CA’s customers and partners are among those working mightily to determine how best to respond, leverage, and capitalize on this fundamental shift toward the cloud model.

And let us all hope that we remember the lessons from previous transformational shifts, including IP networking and the web: Planning, experimentation and prototyping for fast learning, architecture and best practices will enable us to realize the promise of cloud.

We ignore them at our peril...and if we do, we will deserve to be pelted by – meatballs.