At Interop New York last week, I gave a presentation entitled The Democratization of IT. I thought I'd clarify the content of that session, based on some of the feedback on this site. The slides are available on Bitcurrent's website, along with some (free, nonpartisan) research we've done on migrating enterprise IT to the cloud and cloud performance.

The three ages of computing

Humans have been using electronic computers for scarcely seventy years. In the first half of that time, computers were used for two main tasks: counting things (generally, the population) and throwing things at other things (artillery and space missions.)

The computers were gigantic, expensive, and prone to problems; as a result, they were the domain of specialists. Compared to computers, humans were cheap, so we did the "heavy lifting" -- staying up late to maximize their usage, carrying armfuls of punched cards, flipping switches, and coding in assembler.

From the mid-seventies on, the locus of innovation shifted. Client-server architectures and the personal computer took over. We realized that some tasks were better distributed throughout the organization, and millions of small businesses had access to tools that were previously the domain of only large companies. Playing fields were levelled.

Since the late nineties, and the advent of the Web, we've been in a third age of computing. Today, everyone is a technologist, and the computer itself has started a slow, inexorable slide into the background -- what Mark Weiser called Ubiquitous Computing, or Ubicomp, back in 1988.

First: monopoly good

A monopoly occurs when one organization controls a market. It's generally bad for competition and innovation, which is why most countries have something like the Department of Justice to investigate antitrust claims. But there are times when it makes sense to create a legal monopoly -- in particular, when a common resource needs to be created, but isn't attractive in its own rights.

The U.S. did this with the creation of AT&T, which ran a legally sanctioned monopoly in order to build out phone infrastructure. We have other methods to encourage development through exclusivity: railroad corporations were granted land rights; drug companies were given IP protection; and companies like the Hudson's Bay Corporation were given the right to exploit a region. Governments have a monopoly on road building, primarily because it's a job nobody wants.

And until recently, IT had a monopoly on computing. It was costly, complicated, and thankless. It made sense to give IT a monopoly.

Then: monopoly bad

A funny thing happens to monopolies: they tend to work themselves out of a job. In the case of AT&T, other providers wanted access to the local loop, and legislators "unbundled" services that ran over the wires, leading to the eventual breakup of AT&T. Wireless voice and wired data made the market appealing. Innovation soared.

Similarly, as the Hudson's Bay Corporation opened up the Canadian North, others wanted to play. As drug companies' patents expire, they need to find new medicines. Eventually, then, monopolies are bad, stifling innovation and creating inefficient cultures.

IT's days as a totalitarian regime are over

Today, everyone is a technologist. Few of us sleep with our email more than five feet from our heads. If you need proof of this, look no further than your keyboard:  back in the seventies, few workers even knew how to type; today, it's a basic skill. In nearly every company in a developed nation, employees walk in the door with newer technology on their hip than their employer puts on their desk.

Last week, someone asked me what the trigger for the hype around cloud computing was. I answered that it was probably Salesforce.com -- not because of the product itself, but because of how it insinuated itself into large companies. Salesforce completely bypassed the enterprise IT immune system. Sales -- arguably the least technical department in the company -- managed to replace million-dollar, multi-year CRM implementations that IT had worked hard to deploy.

It's not just SaaS, either. Look at the web analytics market. There's no good reason why this isn't a function of enterprise IT -- after all, they had the logfiles needed to extract visitor information from websites. But they were simply too slow to react, so marketing used Javascript to pass visitor information to third-party vendors like Omniture, Webtrends, and Web Side Story.

With cloud computing, this is getting even more obvious. Expense reports used to give IT a sense of security: a line of business employee couldn't possibly procure servers, bandwidth, and data center space because it would be over their departmental limit, so they were forced to go to the IT monopoly for permission -- and generally speaking, had to wait weeks or months.

That's no longer the case. A department can buy computing for $0.10 an hour on Amazon.com, paying as they go. Google App Engine is free to use, up to a point. In one session at Interop, Dan Koffler and Shlomo Swidler showed how to stand up a three-tiered SugarCRM application on Amazon; then scale the app tier vertically; then clone it into a pair of applications, adding a load-balancer; then make a copy of those systems and upgrade them, deleting the old ones; then decommission it.

It took them 40 minutes. And it cost only a few cents.

That's a big change for IT. When computing becomes too cheap to bother expensing, and when the end users -- developers, product managers, and support teams -- are building applications, IT's monopoly has been revoked. I spoke with one telco whose support team, frustrated with the slow pace of deployment by their internal team, built a ticketing system from Intuit Quickbase in a few days. By the time IT realized what was happening, 4,000 people relied on it, and IT had to catch up.

What IT does now

That doesn't mean IT goes away. It means that the CIO has a new role, much as governments have a new role in democracies. It's about governing, aggregating a set of services both inside and out, and making it easy to couple public and private infrastructure. It's about making sure your employees don't hurt themself -- think of the Environmental Protection Agency and the Federal Trade Commission in the US: their job is to make sure that everyone's playing by the rules.

Paula Klein's writeup on my presentation was necessarily brief, and may have made it sound like I was advocating a utopia of social networks and an abandonment of centralized IT. That's not the case; but I do believe that IT needs to readjust its role in the organization in an era where consumers and employees, not companies and organizations, are the locus of innovation.

In the closing session of Interop's Enterprise Cloud Summit, I remarked that "when it comes to cloud computing, the genie is out of the bottle. Enterprises need to stop looking for a cork, and start deciding what to wish for." I think that's a pretty accurate assessment of IT in a world where everyone's a technologist, computers are cheap, and fast, agile business is the norm.

By Robin Bloor

It’s difficult for a CIO today not to be considering a cloud-related strategy. Over the past three years, these hosted services have acquired marketing sparkle and every IT vendor worth its socks has developed offerings.

For large enterprises, two things have become clear: First, Google, Yahoo and other Web-based businesses with very large data centers have demonstrated economies of scale by pointing thousands of servers at a single application. Second, and more important, Amazon has established a thriving and fast-growing business by providing storage and virtual machines on an hourly rental basis.

These two developments have proven that cloud computing has legs, although it is still evolving today. The critical question for the CIO, however, is what to say when the CEO asks, “What are we doing about cloud computing?”

My answer is that the primary motivation for moving systems into the cloud is to reduce costs: staff costs, establishment costs, energy costs and hardware costs. The cloud will not significantly improve the key business processes of your organization, though it will help you develop applications more quickly.

Currently, the major infrastructure-as-a-service (IaaS) options deliver only Intel-based resources running Windows or Linux. And sadly, only about half of all data center applications run on these operating systems. Even with the best intentions, you won’t be moving the other 50 percent to the cloud anytime soon. And some of the Windows or Linux applications aren’t really candidates for IaaS in any case. Anyone using Microsoft Exchange in-house, for example, might move to a hosted service. Call it cloud computing if you like, but in truth that’s business as usual.

Hosted e-mail systems existed long before cloud computing, as did hosted Web sites and other software-as-a-service alternatives to data center applications.

Low-Hanging Fruit
My advice to large organizations, then, is to begin with the low-hanging fruit. But, as for other applications, prove that you can run them in a private cloud first before you turn to an Internet-hosted cloud model. Management will be the key: resource management, application management, performance management, service management and recovery management. A good rule of thumb: If you can’t make it work in your own data center, it’s not going to work in the cloud.

Consider software development in all its aspects. Many companies already supplement this activity with cloud resources because development software is portable and there are obvious benefits. For example, developers no longer have to negotiate with data center staff to get extra resources. It can take days or weeks for a data center to make an extra server available, but in the cloud, you can get one in minutes, pay for it by the hour, and scrap it when it’s no longer needed.

When you look at operational systems, however, you run into more complexity. It’s true that some stand-alone systems can evaporate into the cloud with few consequences, but only if your cloud service provides a management interface and is secure. Most applications are not stand-alone; they have dependencies, and dependencies are not cloud-friendly because they don’t port very well.

The Role of Virtualization
Virtualization, therefore, is the litmus test and should precede any broad adoption of cloud computing. You may also call this server consolidation, or server virtualization. Some view it as creating a private, or internal, cloud.

Yet, virtualization has its own complexities. Anecdotal feedback from large sites that have pursued virtualization projects suggests that diminishing returns nearly always set in once software development and stand-alone systems have been virtualized. Eventually, a point is met where the payoff falls below zero, as management costs escalate and managing the set of virtualized resources becomes increasingly difficult.

All of this must be considered before implementing cloud computing.

[Robin Bloor is President and Chief Analyst, The Bloor Group, and founder, Bloor Research. He is also co-author of the books, Service Oriented Architecture for Dummies, Service Management for Dummies, and Cloud Computing for Dummies. Read his blog here.]

Robin is also a member of Smart Enterprise Exchange and will reply to your comments here.