NEW YORK — Ever since HAL, the self-aware computer, doomed the Discovery space mission in Stanley Kubrick’s 1968 sci-fi classic “2001: A Space Odyssey,” one of the movie’s key themes — man vs. machine — continues to drive debate over how much control we should cede to computers.
As autonomic computing systems that can diagnose problems and then fix them build steam in the 21st century, members of IBM’s research group find themselves answering the question: Are computer scientists creating modern day HALs? Well, to paraphrase one of HAL’s most famous lines: “I’m sorry Dave, I’m afraid they won’t do that.”
What they are doing, said research executives at the first ever International Conference on Autonomic Computing, is freeing up network administrators to focus on larger problems in managing today’s complex networks by infusing self-healing abilities in networks.
“We’re trying to elevate people to a higher level” with autonomic systems work, said Jeffrey Kephart, manager of IBM Research’s agents and emergent phenomena group. Indeed, along with the task of developing networks that are aware of problems and can translate different systems’ problem data, ethnographic research is proceeding apace in order to understand how humans and machines interact, he added.
When asked the HAL question during a press briefing at the conference this week, Steve White, senior manager of autonomic computing for IBM Research, used a camera analogy to explain the autonomic difference. The photo industry once consisted of complex cameras that called for technical know-how in order to get a good picture — how to adjust it for lighting and shutter speed, for example.
“Now, most are point and click,” White added. “Yet none of the manual functions have been taken away.” That’s what is happening with autonomic systems, he added.
Plus, much of the technologies the researchers are advancing with autonomic computing systems are already in use today. This includes Microsoft Word documents that correct spelling as you type, or how Amazon.com mines customer data to make recommendations the next time the customer hits the site, to Web sites that remember your favorite links for you.
Much of the research group’s work involves how to diagnose the problems, and how one network can understand data from different formats that come from problem event logs. “The end goal is homogeneity across heterogenous systems,” White said.
That’s one reason IBM recently submitted a Common Base Event format to OASIS, the technical standards body, as a way to help standardize how event data is read within networks.
In addition, IBM said three companies, Corente, NetFuel and Singlestep Technologies, are slated to produce products that use the Common Base Event format. Partner companies are also integrating IBM’s Autonomic Management Engine (AME) into upcoming
products. The AME, officials explained, monitors events, analyzes them, and then plans and executes
corrections on a computing resource.
The CBE format is one of a number of projects within IBM’s Research group, which is
organized under the company’s software division. For example, the Unity project is researching how
component behaviors, relationships and technologies can support self-management of computing systems. Part of the Unity project includes a prototype of an autonomic data center that configures and heals
Another project, said White, is called Change Management with Planning and Scheduling, or CHAMPS for short. Officials described it as a process by which IT systems are modified to accommodate considerations such as software fixes, hardware upgrades and performance enhancements. CHAMPS is designed to figure out what configuration changes are needed and how such changes would be planned and executed.
IBM’s Event Mining project, meanwhile, is working with how historical data is mined and parsed in order to build signatures for problem diagnosis and problem solving. The goal is to reduce the time between finding a problem in a complex network and fixing it.
“The industry needs a standard to represent problem data,” added David Bartlett, director of business development in IBM’s Autonomic Computing group. “With improved methods of understanding data describing the problem, getting to the root cause of network issues and fixing them can go from weeks to days, even hours.”
And chances are, asking them to open the pod bay doors just wouldn’t compute.