Browse Prior Art Database

System and Method for Java heap usage prediction

IP.com Disclosure Number: IPCOM000235809D
Publication Date: 2014-Mar-25
Document File: 4 page(s) / 70K

Publishing Venue

The IP.com Prior Art Database

Abstract

An analytics mechanism analyses current usage/trends data and determines which time series forecasting method is most applicable for predicting and creates a forecast (e.g. ARIMA, GARCH, Wiener Process etc). If actual usage trends change overtime the analytics mechanism revaluates which predictive forecasting method to use. If required, a more appropriate forecast model can then be used to perform a subsequent java heap forecast. An additional analytics mechanism which will analyze trend in an existing heap dump and determine which objects are the main contributors to java heap usage and also which objects are most likely to cause an upward trend in usage.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

Page 01 of 4

System and Method for Java heap usage prediction


Today as more and more people work for large concurrent distributed systems which run on the Java* platform, there is a need to understand the stability of these systems, typically when multiple concurrent users are connected to a system, more system resources may be used, as more resources are used there needs to be a way to understand how this affects the growth of the Java heap. Typically the Java memory works in a way, that when objects are no longer needed then with periodic memory management and garbage collection, these objects will be released.

However due to a plethora of reasons there may be cases when objects that have been created and used, are not destroyed, similarly there may be cases when an abundance of objects are created but never used, and even with regular housekeeping these objects the memory space is not reclaimed. So with these classes of problems there needs to be a way for administrators of J2EE** systems to be able to determine given current programming standards and given current customer usage how Java heap usage will trend into the future, and if there are trends which may cause system instability then this information should be clearly articulated.

Therefore is it necessary to have a system and method which can do the following:


- Have the ability to predict/forecast Java Heap Usage to a given duration into the future
- Determine which Java objects contribute most to heap growth (by analysis of a heap dump)
- Have the ability to restrict creation of objects which are contributing to excessive growth

As systems become more complex and as there is a move to more distributed concurrent systems there is a need to understand the system health of the systems which provide essential services to users. For example there needs to be a way to understand if a user is running a Cloud based service is there some pattern of behaviour which may cause an increase in Java heap usage, if so at what point into the future will this usage cause a negative impact on a system. Given the dynamic usage of the java heap, a single time series forecasting model is not desirable.

Therefore the inventive steps(s) are as follows:


- An analytics mechanism analyses current usage/trends data and determines which time series forecasting method is most applicable for predicting and creates a forecast (e.g. ARIMA, Wiener Process etc)

- If actual usage trends change overtime the analytics mechanism revaluates which predictive forecasting method to use. If required, a more appropriate forecast model can then be used to perform a subsequent java heap forecast.

- An additional analytics mechanism which will analyze trend in an existing heap dump and d...