Data processing: artificial intelligence – Neural network – Learning method
Reexamination Certificate
1998-03-06
2001-10-30
Davis, George B. (Department: 2762)
Data processing: artificial intelligence
Neural network
Learning method
Reexamination Certificate
active
06311175
ABSTRACT:
BACKGROUND OF THE INVENTION
TECHNICAL FIELD OF THE INVENTION
The present invention relates to complex information technology systems (IT) and, in particular, to continuity analysis techniques for discovering relations among complex events occurring in such systems, and, more particularly, to techniques for improving the performance of such IT systems through iterative system modeling.
BACKGROUND AND OBJECTS OF THE INVENTION
With the exponential growth of the computer and the computer industry, information technology (IT) systems have become increasingly complex and difficult to manage. A typical IT system in even a small company may contain dozens of computers, printers, servers, databases, etc., each component in some way connected to the others across the interlinkage. A simplified example of an interconnected IT system is shown in
FIG. 1
, described in more detail hereinafter.
Although interconnected systems, such as the one shown in
FIG. 1
, offer many advantages to the users, e.g., resource sharing, as such systems grow and the number of component interlinkages increase, the behavior of these complex systems becomes more difficult to predict. Further, system performance begins to lag or becomes inconsistent, even becoming chaotic in nature. The addition or removal of one component, even seemingly minor, could have dramatic consequences on the performance of the whole system. Even an upgrade on one component could adversely affect a distant, seemingly unrelated component. The system and method of the present invention is directed to techniques to better predict the behavior of complex IT systems, offering system administrators the opportunity to identify problem areas such as performance bottlenecks and to correct them prior to a system or component failure.
Conventional approaches to system performance monitoring are inadequate to easily divine the nature of a performance problem in a complex IT system since any data collected in monitoring is generally useless in ascertaining the true nature of the performance difficulty. The system and method of the present invention, however, provide a mechanism whereby system monitoring data is made easily accessible and usable for analyzing current performance and predicting future performance. The present invention facilitates this analysis through use of data mining principles discussed further hereinafter.
In general, data mining is an analysis of data, such as in a database, using tools which determine trends or patterns of event occurrences without knowledge of the meaning of the analyzed data. Such analysis may reveal strategic information that is hidden in vast amounts of data stored in a database. Typically, data mining is used when the quantity of information being analyzed is very large, when variables of interest are influenced by complicated relations to other variables, when the importance of a given variable varies with its own value, or when the importance of variables vary with respect to time. In situations such as these, traditional statistical analysis techniques and common database management systems may fail or become unduly cumbersome, such as may occur when analyzing an IT system.
Every year, companies compile large volumes of information in databases, thereby further straining the capabilities of traditional data analysis techniques. These increasingly growing databases contain valuable information on many facets of the companies' business operations, including trend information which may only be gleaned by a critical analysis of key data interspersed across the database(s). Unfortunately, because of the sheer volume and/or complexity of the available information, such trend information is typically lost as it becomes unrecoverable by manual interpretation methods or traditional information management systems. The principles of data mining, however, may be employed as a tool to discover hidden trend information buried within the pile of total information available.
Such data mining techniques are being increasingly utilized in a number of diverse fields, including banking, marketing, biomedical applications and other industries. Insurance companies and banks have used data mining for risk analysis, for example, using data mining methods in investigating its own claims databases for relations between client characteristics and corresponding claims. Insurance companies have obvious interest in the characteristics of their policy holders, particularly those exhibiting risky or otherwise inappropriate activities or behaviors adverse to the companies' interests, and with such analyses, are able to determine risk-profiles and adjust premiums commensurate with the determined risk.
Data mining has also found great success in direct marketing strategies. Direct marketing firms are able to determine relationships between personal attributes, such as age, gender, locality, income, and the likelihood that a person will respond to, for instance, a particular direct mailing. These relationships may then be used to direct mailing towards persons with the greatest probability of responding, thus enhancing the companies' prospects and potential profits. Future mailings could be directed towards families fitting a particular response profile, a process which could be repeated indefinitely and behaviors noted. In this sense, the data mining analysis learns from each repeated result, predicting the behavior of customers based on historical analysis of their behavior.
In the same manner demonstrated hereinabove, data mining may also be employed in predicting the behavior of the components of a complex information technology (IT) system, such as the one shown in
FIG. 1
or a more complicated one found in the business environment. Similar approaches as above with appropriate modifications can be used to determine how the various interconnected components influence each other, uncovering complex relations that exist throughout the IT system.
As discussed, multiple applications will be operated within a common IT infrastructure, such as the one shown in FIG.
1
. Often, these applications will utilize some of the same resources. It is obvious that the sharing of IT infrastructure resources among different applications may cause unexpected interactions on system behavior, and that often such unexpected interactions, being non-synergistic, are undesirable. An example would be multiple business applications sharing a router within an IT system. As illustrated, a particular application, e.g., an E-mail service, burdens a router in such a way that other applications do not function well. In this example, it is reasonable to expect numerous applications to, at times, share usage of the router. Traditional systems management techniques may prove difficult in determining which specific application is causing loss of system performance. This example further explains why there is a need to find hidden relationships among IT system components and applications running in such environments. By way of solving the problem in this example, it may be necessary to reroute E-mail traffic through another router to obtain adequate performance for the other applications.
Traditional IT system management is now generally defined as including all the tasks that have to be performed to ensure the capability of the IT infrastructure of an organization to meet user requirements. Shown in
FIG. 2
is a traditional IT systems management model, generally designated by the reference numeral
200
. Essentially, there are groups of system administrators
210
having knowledge of the IT infrastructure, such as the one shown in FIG.
1
and generally designated herein by the reference numeral
220
, which they are managing. Typically, the knowledge of the infrastructure
220
is scattered among the various personnel making up the system administrator group
210
. The total of this knowledge is limited to the sum of the individual administrators' knowledge, where invariably there is a great deal of redundancy of knowledge. This redundancy may be
Adriaans Pieter Willem
Gathier Marc
Knobbe Arno Jan
Davis George B.
Holmes Michael B.
Jenkens & Gilchrist PC
Perot Systems Corp.
LandOfFree
System and method for generating performance models of... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System and method for generating performance models of..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System and method for generating performance models of... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2610444