Systems that can manage and "heal" themselves with minimal human intervention will likely become reality. But it...
will be at least a decade before the complex maze of products required for true "self-healing systems" are fully up to the task, and it will take even more time for the technology to reach the majority of users in corporate America.
The concept is still new enough that it's being called different things. IBM Corp. coined the term "autonomic computing" to describe this notion of computers regulating themselves, so-named because of its intended similarities to the human autonomic nervous system. Forrester Research Inc., in Cambridge, Mass., calls it "organic" computing. Hewlett-Packard Co. terms it an "adaptive infrastructure."
Frank Gillett, a principal analyst at Forrester, says that the goal of all these projects is to hide complexity and to make systems simpler to operate. The way self-healing systems work is similar, in some regards, to the workings of a thermostat, which regulates several subsystems to make the temperature inside a home comfortable. "It's a really simple interface that uses technology to automate [subsystems,] for me," Gillett said.
Vendors are selling self-managing systems to customers by emphasizing the savings buyers can accrue in terms of more efficient use of resources, increased productivity in the data center and, ultimately, reduced staff costs, since fewer employees will be needed to run things.
But the advent of these "smart" systems is still a long way off. Before anyone can talk about personnel cutbacks, a whole mess of related technology needs to be perfected, including virtualization, automated provisioning, grid computing, on-demand computing and Web services. These are all various pieces of different vendors' visions for self-managing systems.
Because of the complexity of the task, some wonder whether it will, indeed, ever come to pass. "Will it happen? Maybe. But if it does, it will be much more difficult than any of the participants anticipate," said David Boyes, president and CTO at Sine Nomine Associates, a consultancy in Ashburn, Va. "The real challenge will be getting the industry to agree on a common set of interfaces and functions" to be able to measure what's going on within and across a data center, even before anything can begin to "heal" itself.
Jon Oltsik, founder of Hype-Free Consulting, in Acton, Mass., says that there's a common perception that, to be successful, self-managing systems must work across multiple vendors' platforms. "This is much more PR value than it is reality, and [it] will take years to roll out," he said. "If you have to work with three different models for your major computer systems, that's a bad thing."
Boyes also points out that adding any major new capabilities to products will likely make them more expensive. But this higher price tag might well be offset by reduced operational costs.
A more serious issue is that it will take a very long time for customers to replace their existing infrastructures with smart systems. There will be a years-long transition period, during which enterprises will use a mixture of old and new technology. The old will not be managed in the same way as the new.
Still, Boyes allows that if self-managing systems do catch on, they could bring some major benefits. "More than two-thirds of the costs in a data center are for people and operational tools," he said. The aim should be to reduce infrastructure so you don't require as many computing resources, he added.
There are already some smart products and services starting to reach the market today from Hewlett-Packard, IBM, Sun Microsystems Inc. and some smaller vendors. Additionally, in mid-March, Microsoft announced its Microsoft System Center, which promises end-to-end application and system management.
Forrester's Gillett says that customers who get started on this path early stand to save, over a five-year period, up to 50% in expenditures on services and new products. Specific savings, of course, will vary by company, he said.
Any savings, however, would be based on a company's ability to corral the four major systems in a data center, including servers, storage, software and wide area networks. Implementing storage management products alone, for instance, may yield some savings, but not in the 50% range, Gillett cautions.
When all is said and done, few observers expect self-managing computers to mean widespread job cuts in the data center. Sure, the lowest-level operator jobs will probably go away, but that is happening anyway, experts say. Instead, a more likely scenario is that, as one set of problems disappears, a whole new class emerges -- and people will still be needed to help solve them.
"Even if the efficiency of the data center improves, there are always going to be new initiatives, like wireless," Oltsik said. "There will be new models that aren't really thought through yet, so you're really just moving the problem."
>> Next time: A look at what specific vendors are doing in this area.
For More Information: