Neural Agents

Comment: Several years ago, I was the leader of an operationalized telecommunication cell. The purpose of the cell was to monitor the effectiveness and readiness of the telecommunications in support of the ongoing operations. The staff regularly turned over due to the operational tempo and I had to train new staff quickly. I did so by preparing a series of technical briefs on topics the cell dealt with. This brief was dealing with  Neural Agents which I have updated and provided additional postings that paint a picture of potential advanced systems. 

Neural Agents
by
JT Bogden, PMP

Figure 1: Agent Smith
The Matrix movie franchise

Neural Agents have this spooky air about themselves as though they are sentient and have clandestine purpose. The movie franchise the 'The Matrix', Figure 1, made use of Agent Smith as a artificial intelligence designed to eliminate zombie processes in the simulation and human simulations that became rogue such as Neo and Morpheus.  In the end, Agent Smith is given freedom that results in him becoming rogue and rebellious attempting to acquire increasing powers over the simulation. 

The notion of artificial intelligence has been around forever. Hollywood began capturing this idea in epic battles between man and machines in the early days of Sci-Fi.  More recently, the movie "AI" highlighted a future where intelligent machines survive humans. Meanwhile, the Star Trek franchise advances intelligent ships using biological processing and has a race of humanoid machines called the Borg.  Given all the variations of neural technologies, the Neural Agent remains a promising technology emerging in the area of event monitoring but not acting quite as provocative as Agent Smith. The latest development in neural agents in support of artificial intelligence. Neural agents, Neugents which are not related to Ted Nugent, are becoming popular in some enterprise networks. 

Companies can optimize their business and improve their analytical support capabilities as this technology enables a new generation of business applications that can not only analyze conditions in business markets, but they can also predict future conditions and suggest courses of action to take. 

Inside the Neugent

Neural agents are small units or agents, containing hardware and software, that are networked.  Each agent has  processors and contain a small amount of local memory. Communications channels (connections) between the units carry data that is encoded usually on independent low bandwidth telemetry. These units operate solely on their local data and input is received from over the connections to other agents. They transmit their processed information over telemetry to central monitoring software or other agents. 

The idea for neugents came from the desire to produce artificial systems capable of “intelligent” computations similar to those of the human brain. Like the human brain, neugents “learn” by example or observations. For example, a child recognizes colors by examples of colors. Neugents work in a similar way: They learn by observation. By going through this self-learning process, neugents can acquire more knowledge than any expert in a field is capable of achieving.

Neugents improve the effectiveness of managing large environments by detecting complex or unseen patterns in data. They analyze the system for availability and performance. By doing this, neugents can accurately “predict” the likelihood of a problem and even develop enough confidence over time that it will happen. Once a neugent has “learned” the system’s history, it can make its predictions based on the analysis, and it will generate an alert, such as: “There is a 90% chance the system will experience a paging file error in the next 30 minutes”.

How Neugents Differ From Older Agents

Conventional or older agent technology requires someone to work out a step-by-step solution to a problem then code the solution. Neugents, on the other hand, are designed to understand and see patterns, to train. The logic behind the neugent is not discrete but instead symbolic.  They assume responsibility for learning then adapt or program themselves to the situation and even self-organize. This process of adaptive learning increases the neugent's knowledge, enabling it to more accurately predict future system problems and even suggest changes.  While these claims sound far reaching, progress has been made in many areas improving adaptive systems. 

Neugents get more powerful as you use them. The more data it collects, the more it learns. The more it learns, the more accurate its predictions. This solution comes from two complimentary technologies: the ability to perform multi-dimensional pattern recognition based on performance data and the power to monitor the IT environment from an end-to-end business perspective.

Systems Use of Neugents and Benefits

Genuine enterprise management is built on a foundation of sophisticated monitoring. Neugents apply to all areas. They can automatically generate lists for new services and products, determine unusual risks and fraudulent practices, and predict future demand for products, which enable businesses to produce the right amount of inventory at the right time. Neugents help reduce the complexity of the Information Technology (IT) infrastructure and applications by providing predictive capabilities and capacities. The logic behind the neugent is not discrete but instead symbolic. 

Neugents have already made an impact on the operations of lots of Windows Server users who have already tested the technology. They can take two weeks of data, and in a few minutes, train the neural network. Neugents can detect if something’s wrong. They have become a ground-breaking solution that will empower IT to deliver service that today’s digital enterprises require.

With business applications becoming more complex and mission-critical, the us of neugents is more necessary to predict then address performance and availability problems before downtime occurs. By providing true problem prevention, Neugents offer the ability to avoid the significant costs associated with downtime and poor performance. Neugents encapsulate performance data and compare it to previously observed profiles. Using parallel pattern matching and data modeling algorithms, the profiles are compared to identify deviations and calculate the probability of a system problem.

Conclusion

Early prediction and detection of critical system states provide administrators an invaluable tool to manage even the most complex systems. By predicting system failures before they happen, organizations can ensure optimal availability. Early predictions can help increase revenue-generating activities as well as minimizing the associated costs due to system downtime. Neugents alleviate the need to manually write policies to monitor these devices.

Neugents provide the best price/performance for managing large and complex systems. Organizations have discovered that defining an endless variety of event types can be exhausting, expensive and difficult to fix. By providing predictive management, Neugents help achieve application service levels by anticipating problems and avoiding unmanageable alarm traffic as well as onerous policy administration.