Artificial Intelligence and Enterprise Management

By Julie Craig

(Back to article)

Interest in artificial intelligence (AI) as it relates to IT seems to be growing. Back in December, Network World published an article titled National Science Foundation commands artificial intelligence revolution. Apparently, NSF funds AI research to the tune of $90 million annually as part of its Information and Intelligent Systems (IIS) grant program. Researchers can apply for as little as $500,000 or as much as $3 million for projects related to AI, robotics, machine learning, and other key areas.


In January, Computerworld followed suit with an article entitled Future Watch: A.I. comes of age. The article includes some fascinating information about robotics and machine learning. What is most interesting to me is it describes AI as able to real-world problems by analyzing complex input and learning to adapt accordingly. I was intrigued because, while AI is a somewhat esoteric subject in the non-IT world, yeas ago enterprise management products were among the first AI products to become commercially viable.


In the Beginning


Starting with the original network management tools made possible by the standardization of simple network management protocol (SNMP) in the 1980’s, up to and including today’s sophisticated service oriented architecture (SOA) products, management solutions were among the earliest commercial offerings to productize AI. Admittedly, early products were primarily static systems programmed to display errors based on SNMP traps, and, at most, to perform Y in response to X. Very soon, however, they incorporated correlation capabilities with the innate intelligence to isolate the most likely cause of a network disruption and to notify support specialists about problems.


Today’s solutions have evolved to the point where they feature much higher levels of machine learning and adaptation. Particularly in the past two years, evolution seems to have accelerated and many of today’s products are capable of supplementing the expertise of in-house technology specialists very effectively. I discussed some such products and capabilities in last month’s article on predictive analytics. What does this mean for the CIO? Products with high levels of embedded expertise offer significant opportunities for cost efficiencies.


Products with adaptive capabilities “learn” the normal characteristics of an IT environment by “observing” infrastructure and applications during execution, and collecting and storing data. Often as granular as time of day, day of the week, and time of the month; this data is utilized to analyze real-time execution in context to historical trends. Based on complex algorithms that are often patented, some solutions can maintain rolling baselines, predict application issues before they become full-blown problems, or report whether an infrastructure change has made things better or worse.


Tomorrow's AI


As higher levels of intelligence become embedded in enterprise management solutions over time, we will be able to offload a growing percentage of application management tasks to automation. In fact, as I talk with vendors about products and their capabilities, the picture I have in the back of my mind is of an autonomic computing system. Going under different names depending on the vendor, we are seeing autonomic technologies coming from IBM and others, Dynamic Systems from Microsoft, and the Adaptive Enterprise coming from HP. These terms refer to systems—and by systems I mean collections of technology, not stand-alone devices—that incorporate AI for self-management or to manage other systems, to self-configure or self-heal, or to perform other functions incorporating high levels of expertise.


Taking this trend to its logical end state, systems and application management will become the domain of engineers embedding expertise in software, rather than systems administrators sitting at a console. While IT will always require hands and heads, it’s likely the effort and time that goes into support will diminish over time; with higher percentages of IT budget and headcount allocated to new projects and products.


This, of course, is the H


oly Grail of enterprise management and will require continued evolution on the part of enterprise management vendors. This vision requires management systems that “watch” technology ecosystems, “understand” normal functioning, “analyze” real-time performance against norms, then “act” to automatically solve known problems. Enterprise management will align more closely with AI over time, as autonomic systems increasingly take on more of the characteristics of human intelligence.


We are already starting to see some of these capabilities from multiple vendors, including many of those mentioned in last month’s article. Most suite vendors, and some smaller ones, already offer products capable of automatically provisioning virtual servers in response to application-related problems. A related “hot off the presses” technology recently announced by RNA Networks is “virtual memory”. RNA delivers “virtual memory” pools across multiple clustered servers. Combined with automated provisioning of virtual servers, capabilities such as these will enable true “elastic computing” by eliminating one of the biggest risks of virtual server deployment—over utilization of host system resources.


Intelligent products can be excellent investments, as they supplement in-house expertise with world-class technology expertise from some of the best minds in the business. Encapsulated in software and productized, I’m convinced that such products will help drive down IT support and administration costs over time. The fully autonomic systems that I envision, however, will be years in the making. And while it might not be fair to compare products to a Holy Grail that is likely years in the future, this vision does provide a yardstick for measuring the evolution of management products over time.


Today, such products already deliver substantial reductions in the amount of time required to detect, analyze, and solve application-related problems. As time goes on these products will continue to evolve and cost efficiencies will become more marked.


Julie Craig is a senior analyst with Boulder, Colo.-based Enterprise Management Associates, an industry research firm focused on IT management. Julie can reached at jcraig@enterprisemanagement.com. Additional research into autonomic computing is available at EMA’s site at www.emausa.com. In addition, EMA is in the process of creating a new, End-to-End Application Management Online Guide, which will be available by the end of Q1, 2009. Continue to check EMA’s site for this Guide, which will feature detailed profiles of multiple application management products and be free of charge.