Browse Prior Art Database

Method for Predictively Elevating Heterogenious System Monitoring in Regards to Capturing Complex Events

IP.com Disclosure Number: IPCOM000216781D
Publication Date: 2012-Apr-19
Document File: 7 page(s) / 153K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a method to intelligently identify the source of problems within a heterogeneous software product either within the development and test cycle or in the field. The invention uses machine learning coupled with human intervention to intelligently capture interesting events in the complex system via the use of approximate pattern matching of system components leading up to and at the time of the fault/event of interest.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 24% of the total text.

Page 01 of 7

Method for Predictively Elevating Heterogenious System Monitoring in Regards to Capturing Complex Events

This invention proposes to solve the issue of how to intelligently identify the source of problems within a heterogeneous software product either within the development and test cycle or in the field.

Analysts cannot always fully evaluate the complexities of the interactions or implications of concurrent execution of loosely coupled software components within a complex software system during the scope of CVT/SVT software testing. The permutations necessary to expose interesting events or faults in the system may be unknown or too numerous to approach. This heuristically addresses this problem by using machine learning coupled with human intervention to intelligently capture "interesting events" in the complex system via the use of approximate pattern matching of system components leading up to and at the time of the fault/event of interest.

The intention of this invention is to capture relevant data about a set of events that may happen in a complex software system while affecting the system to a smaller degree than might otherwise happen.

An example of a problem that is typically difficult to expose is a memory leak. Perhaps only a specific code path or pattern of execution of a software system is required to surface a memory leak within a certain component. However, the exact nature of the prerequisites is unknown ahead of time by development or the software test team. Usually, within the scope of a test plan for a specific release, testers plan/execute only a subset of all possible test-cases/scenarios during a release cycle. This subset might fail to bring about the concerning behavior. Alternatively, in the field, the interesting system fault might intermittently occur for a customer for unknown reasons (termed a test-escape).

The difficulty of the event and time required to understand and evaluate the customer's issue may be time-prohibitive or too pervasive (re: the customer's process and data) for the customer's satisfaction, as it may take multiple iterations to capture meaningful information.

An alternative methodology for exposing rarely encountered system faults is to provide debug builds of either the components or the overall system that testers must apply. This package might feature break points in the code or alternative code implementations to isolate the code path of interest. This can also be invasive for the customer and can impact the "uptime" or performance of the system - something that may not be desirable or tolerable for the customer.

The primary drawbacks of the usual methods for isolating or evaluating complex issues in a heterogeneous software system are customer impacts to performance, uptime, convenience, and timeliness for providing a solution. The drawbacks of employing the usual subset of test scenarios is that interesting events may not be exposed or

1


Page 02 of 7

understood in testing, lead...