Managing IT resources with an agent

By Zatni Arbi

JAKARTA (JP): Imagine yourself working for a bank with hundreds of branches scattered all over Indonesia as the manager responsible for its huge, distributed network. How can you make sure that all your company's IT resources are well-maintained and that their capacity is always adequate?

With today's computer hardware and software technologies, you won't have to physically travel around the country and visit each of the branches to check each individual PC, server, mini and even mainframe computer that it may have. Nor will you have to hire highly skilled -- and highly paid -- technical support personnel to man the help desks. A type of software suite called "enterprise management framework" will do it for you. Currently two strong contending products are fiercely battling it out in this field: Tivoli Enterprise from Tivoli (an IBM company) and Unicenter TNG (TNG stands for 'The Next Generation') from Computer Associates.

These suites of end-to-end management tools promise to help large corporations manage their widely dispersed IT resources through a centralized control system by automating a lot of functions, including maintenance and asset management. As you can imagine, in such a huge environment there may be UNIX servers, NT servers, OS/2 servers, a Novell network operating system and many other platforms. Because of such diversity, these "umbrella" tools are supposed to be able to also work on top of all these different platforms and facilitate management so that the utilization of the resources can be maximized and disruption can be minimized.

Two typical problems that IT personnel in such huge computing environments usually face are downed servers and an unexpected jump in the number of users that means unanticipated increased demands for resources. Unlike our home PC, which we can replace at any time if it crashes, the servers in these organizations cannot stop functioning just because maximum capacity has been overstepped or one component has failed. Just imagine how frustrated we will become each time we are told by the teller that the bank's server is down and we cannot make any transaction.

Both Tivoli Enterprise and Unicenter TNG use the so-called 'agents' that are deployed to the individual resource in the enterprise network, including the PCs and the notebooks. These software agents regularly collect data from their assigned location and report any extraordinary event in their target environment to the central management software. Based on these reports, the IT personnel, who may be notified through pagers or E-mail in case the light has turned from green to yellow, will then be able to manage the network better by reallocating resources as needed, take preventive action or pinpoint the source of a problem quickly when it occurs.

The increasing importance of complete management of IT resources has led companies such as Computer Associates (CA) to find a new way of simplifying the immensely complex task. One reasonable option is to use a neural network to predict when capacities will be used up to their maximum so that actions can be taken to prevent disasters.

Neural Agent

You may still recall the technology called neural networks. Simply put, a neural network is an attempt to imitate the sophistication and intelligence of the human brain. It is called a network because it consists of numerous "units" of simple processors--each with its own memory--that are connected to each other to form a system that tries to mimic the brain.

The most important aspect of the resemblance is the ability to sort out patterns and learn from trial and error. With it comes the ability to discern and extract the relationships underlying the data that we feed it with. With these two abilities, the artificial intelligence system can even make decisions based on a set of dynamic rules. In addition, the more data it has the more intelligent a neural network will become.

As an application of the neural network, the new Neugents from CA also learns as it goes. The name Neugents actually comes from 'Neural' and 'Agent'. This application is actually a part of CA's strategy in moving to its next version of Unicenter TNG, i.e., Unicenter TND (D for 'Dimension').

What, specifically, does Neugents do to help simplify the management of an enterprise IT infrastructure? Its agents, which are deployed to mission critical servers, will continuously monitor a number of parameters in the systems. Based on these parameters and with the help of the historical data that has been collected at the Unicenter TNG, Neugents will then be able to make a predictive statement such as "There is a 70 percent probability that Server X will crash in an hour."

The server crash may be caused by overloading, for example, as the number of its users keeps rising while its capacity remains constant. When the IT administrator sees such a message in his Unicenter console, he will then be able to make a proper decision to prevent the crash, for example by temporarily shutting down other services and allocating the freed resources to accommodate the increasing number of users at server X.

Can a human being do what Neugents does? As Kieandy Susanto from CA reminded me, human beings can only process information in a limited number of dimensions, perhaps three of them at the most. A computer server, on the other hand, has tens of parameters that determine its overall performance, including memory, hard disk, I/O and processing capacities. It will be very hard for a human brain to process all these data and make a quick decision based on them. A computer, on the contrary, has no problem handling multidimensional data. Herein lies the benefit of Neugents. It can process data in dozens of different dimensions and make a prediction based on them.

Interestingly, although we may not realize it, the neural network-based systems and parent technologies (artificial intelligence and expert systems) are already widely in use all over the world although they have long ceased to be buzzwords. The latest generation of optical character recognition and voice recognition tools already incorporates the neural network technology, and so do the systems that are used by banks to estimate investment risks, for example. However, CA has been the first to bring the technology to the enterprise management tools, and it may be some time before other vendors will follow suit.