IBM calls it “autonomic computing.” Hewlett-Packard labels it the “adaptive enterprise,” and Microsoft has its “dynamic systems initiative.” Sun Microsystems offers “N1,” while Hitachi touts “harmonious computing.” Forrester Research calls all of these efforts “organic IT.”  

No matter how it’s pitched by the big boys of technology, the vision of self-healing computers—networks that monitor themselves, sense problems, and fix them automatically—is quite far from today’s reality.

The ultimate goal is to remove humans from the equation and have computers fix themselves. These so-called autonomic or self-healing systems could, for example, isolate a faulty line of code, retrieve a patch for it, and then fine-tune the application for optimal performance. IT administrators would be liberated from the tedium of system maintenance and repair, and self-configuring, self-optimizing computers and networks would automatically run at peak performance all the time.

But keeping networks and applications up and running still causes major headaches for companies with two complaints—not enough technicians and staggering costs. Just ask Victor Kellan, who has spent most of the last 15 years looking for a network management product that lives up to its name. According to Mr. Kellan, many of the problems that occur on a network result from the interplay among devices and software, but that interaction remains an enormous blind spot to all of these “intelligent” management products.

It’s like a roomful of relatives all talking about the same set of problems at the same time, but in different languages with no one listening, and no structure for finding solutions to their common afflictions.

“We were getting flooded with information, with no filtering, no ability to provide any proactive services and no ability to have any kind of escalation procedures or response criteria,” says Mr. Kellan. Corporations are expected to spend more than $21 billion worldwide next year on managing their applications, according to IDC.

Solving this problem was a full-time occupation for Mr. Kellan, the founder and president of LAN Solutions, a network administration company based in McLean, Virginia. Mr. Kellan’s company monitors and manages IT resources for companies in the Washington, D.C., area. A year ago, Mr. Kellan finally found what he was looking for in a product called Unity from Seattle’s SingleStep Technologies.

Unity did something that none of the other products—even the very expensive ones— could do. It extracted massive amounts of performance information from the existing device and systems managers on the network, collected it in a common database, and manipulated the data for clues as to the health of the entire network. It was like getting all the chattering neurotic relatives to take turns talking to a group therapist in a common language.

SingleStep’s Unity is a software suite based on an IBM engine that uses adapters to extract management information from the enterprise network. That information becomes the raw material that forms the basis of an emerging technology IBM calls autonomic computing.

Last year IBM began distributing a free “autonomic computing toolkit” that helps its partner companies build products that fit into IBM’s emerging autonomic computing architecture. IBM already claims to sell 50 different hardware and software products that offer a total of more than 400 features with autonomic computing capabilities.

“Suddenly we had the material to automate responses to errors. If an engineer has a systematic way of fixing a problem, we could automate that process,” says Mr. Kellan. “There could be three or four repair options before the most expensive one of dispatching a human to fix it.”

Automating the Future

Like most grand technology visions, this one has always been a little out of focus. Despite the visions of self-healing systems spun out by big technology players, the industry has seen little evidence that the vision is becoming reality.

But a new crop of startup companies is also starting to make advances. If successful, they’ll save corporations perhaps billions of dollars in IT repair, maintenance, and staffing costs, not to mention countless headaches. And in turn, they may spark a new wave of mergers, acquisitions, and IPOs.  

SingleStep, with its Unity product, is one of this new breed. CEO Chris Noble says the company’s network management system automatically analyzes information, checks network settings based on policies, and performs certain tasks, such as shutting down access points if necessary. “We’re automating what a network operator would be doing,” says Mr. Noble. SingleStep, which has partnered with IBM, raised $11 million in venture capital, including $5.3 million in a Series B round in March 2004.

“There’s a good possibility some of the first movers will be acquired,” says Bob Locklear, managing director of Agave Capital, which invested in SingleStep. “As the technology matures, eventually some companies may have IPOs.”

Healing Visions

While large technology vendors such as IBM will supply the engines underneath any future autonomic system, smaller companies such as SingleStep will develop supporting technology and customization around it. Venture capital investors, like Tom Clancy, managing director of Enterprise Partners Venture Capital (EPVC), see opportunity in those smaller companies. “It’s a very complex and big company game,” says Mr. Clancy. “You can make more targeted investments in companies with autonomic computing as a theme.”

EPVC’s investments include Breach Security, which closed a $7.5-million Series A funding round in August. The Carlsbad, California, Internet security firm is developing a technology that automatically learns the profiles of attackers and adapts its defenses accordingly. Netsift, another early-stage company backed by EPVC, touts an Internet security system that automatically generates unique signatures used in combating worm attacks.

Symbium, a two-year-old Ottawa-based company, is developing autonomic technology that monitors and corrects Windows-based servers. In the event of a worm attack, for example, the technology will extract the intruder and automatically reboot the system. According to Pat DiPietro, managing general partner at VenGrowth Capital Partners, which invested in Symbium, the firm’s autonomic technology will analyze a “dead” server, determine what’s wrong, fix the problem, and bring the system back to life automatically. Last June Symbium announced it raised $6.75 million in first-round venture funding, for a total of $7.75 million, including a round of angel investors.

Pipe Dreams

Broadband networks are expected to play an important role in deploying self-healing technologies in lieu of sending technicians to perform onsite repairs. For example, SupportSoft’s technology helps corporations manage IT systems by automating the updating, management, and service of software.

Cadir Lee, SupportSoft co-founder and vice president of engineering, estimates that 60 percent of the company’s revenue now comes from autonomic computing. For the nine months ending September 30, SupportSoft, based in Redwood City, California, posted total revenue of $46 million. Mr. Lee envisions such technology will play an important role in managing a wide range of devices, including digital televisions, PDAs, smart phones, and home networks.

At Network Physics in Mountain View, California, CEO David Jones sees opportunity in one area of autonomic computing. Founded in 1999, Network Physics developed a network-sensing appliance that can collect and analyze performance and service data across a complex, enterprise-wide network. A few months ago, the company began selling a version of its device that’s compatible with IBM’s autonomic computing architecture.

Mr. Jones believes such devices eventually will become essential in helping autonomic systems function. In September, Network Physics announced it raised $13 million in a third round of funding led by Trinity Ventures, bringing its total funding to $24 million. “Autonomic is a big word, but it’s not an all or nothing proposition,” says Mr. Jones. “There will be phases of adoption.”

The first phase, as with most new technologies, will involve some skepticism and resistance. Corporate IT executives may not be willing to trust a machine or a technology they can’t see to make decisions for them. After all, the idea of leaving control of fixing and maintaining expensive IT infrastructure in the grasp of an automated computer program makes some CIOs shudder.

“Self-discover, self-manage, and self-healing scares the hell out of customers,” says Mark Sigal, chief executive of UXComm. Mr. Sigal’s company, based in Beaverton, Oregon, sells what it calls an adaptive management system for Intel-based servers. UXComm touts autonomic computing technology as an important feature in its product.

But if the technology is truly self-healing, it’s hard to argue with something that could automate tasks now performed by expensive IT staff.

Big Company Maneuvers

Meanwhile, big computer manufacturers are jockeying for position among self-healing technologies. In addition to adding new self-correcting features to its products, Microsoft continues to lead an industry-wide campaign to get hardware manufacturers and software developers to follow standards, such as XML, and Windows architecture guidelines to promote compatibility—and to drive down the costs of maintenance and ownership of its own products. Microsoft wants to reverse a trend where corporations spend more on maintaining existing systems than on buying new ones.

HP is firing its own salvo with new products for measuring system and network performance. The company will engineer its network management solutions to react to those metrics. The new systems and software products are modular, so they are easier to deploy and self-adjust to changing workloads.

But products based on autonomic capabilities are only as good as their ability to access the millions of bits of information that the average network generates. Many vendors use proprietary formats for storing and manipulating that information. So two years ago, IBM and Cisco submitted a specification to the OASIS standards body, called the Common Base Event format, which they hope will become a standard for accessing, analyzing, and fixing problems using self-healing technologies.

With the emergence of new buzzwords like autonomic computing, a certain caveat emptor must be considered, says Karsten Schwann, professor and director of the Center for Experimental Research at Georgia Institute of Technology. A battery of products already exist—network devices, computer systems, and software—that self-manage to some degree. But that’s a lot different from “autonomic,” although some manufacturers may make that claim.

“Some systems already are self-managing, so many people can claim they are delivering systems that are autonomic,” says Mr. Schwann.

For true autonomic computing to become a reality, the technology industry must work together to build products that fit together, communicate with each other, and collaborate to find and fix problems. “Today you’ve got pieces,” says Audrey Rasmussen, vice president of Enterprise Management Associates, an IT consulting firm in Boulder, Colorado. “Going across the infrastructure is where the groundbreaking will be.”

Future Remedies

More hurdles lie ahead. For example, who decides which applications shut down and which remain running whenever problems arise? What databases and servers take priority? Asking those questions will surely prompt conflicts between employees and customers, business partners, corporate departments, and executives and staff.

In its position papers, IBM cites how the technology, in order to progress, will need the input of psychologists, “human factor researchers,” mathematicians, accountants, economists, scientists, and others to sort out the tangle of organizational, economic, and commercial priorities.

“This is as challenging as it gets,” says Alan Ganek, IBM’s vice president of autonomic computing. “Improving individual products alone won’t get you over the goal line.”

According to venture capitalist Mr. Clancy, “There are hundreds of millions of addresses connected to the Internet, going to billions and perhaps trillions…. Human intervention will not allow us to scale to that level. For humans to go in and intervene manually won’t work.”

Armando Fox, assistant professor of computer science at StanfordUniversity, adds: “Our ability to build complex systems is much better than our ability to fix them.” At Stanford, Mr. Fox is exploring how rebooting systems, which he considers a basic autonomic technique, may play a role in an autonomic computing milieu. The downside of rebooting can mean loss of data, corrupt files, and downtime. Stanford researchers are considering the possibility of “micro” rebooting—resetting part of a system while continuing to run other operations.  

Meanwhile, the complexity of IT continues to grow. Given the seemingly infinite problems that could occur when different hardware, software, applications, users, and other elements intersect, opportunity certainly abounds for any company with a true solution.

.