Browse Prior Art Database

Live Garbage Collection Tuning Recommendations

IP.com Disclosure Number: IPCOM000180171D
Original Publication Date: 2009-Mar-05
Included in the Prior Art Database: 2009-Mar-05
Document File: 3 page(s) / 76K

Publishing Venue

IBM

Abstract

This disclosure describes a system for providing live analysis of garbage collection activity in order to provide continuous feedback and recommendations. The novelty is the combination of live monitoring with sophisticated analysis and recommendations, usually only found in offline tools.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 50% of the total text.

Page 1 of 3

Live Garbage Collection Tuning Recommendations

Garbage collection is critical to the performance of Java* applications. When badly tuned, it can divert CPU resources away from the application - when very badly tuned it can even cause application crashes with OutOfMemoryErrors. In contrast, when well tuned it can provide a significant performance boost. Garbage collection behaviour can also provide useful insight into application behaviour and provide a lightweight way of detecting memory leaks or excessive memory requirements (footprint bloat).

    Verbose GC allows garbage collection behaviour to be visualised and can provide input for analysis tools which provide recommendations. Such tools, however, are intended for offline analysis and do not provide recommendations dynamically. They are therefore not very useful for pre-empting out of memory crashes or detecting excessive GC overhead on the fly. Other tools allow live visualisation of GC statistics but they do not offer unskilled users any guidance to interpreting the visualisations.

Prior Art

    Sun provide the GC portal with tuning recommendations ( http://

     ava.sun.com/developer/technicalArticles/Programming/GCPortal/), but they explicitly say "The Portal does not provide dynamic recommendations or automatic tuning. ". Our invention covers dynamic recommendations.

    BEA's Mission Control provides live visualisations of garbage collection data but there does not appear to be any analysis text or recommendations. Our invention covers dynamic recommendations.

    IBM**'s GC and Memory Visualizer provides garbage collection recommendations but can only run off-line, and only against verbose gc output. Our invention covers live recommendations.

    Several tools provide visualisation of live garbage collection statistics. Such tools include Sun's JConsole (for a strictly limited number of GC metrics), Sun's VisualVM, and BEA's Mission control. All of these tools, however, rely on the user having a good understanding of garbage collection tuning and do not provide any suggestions or recommendations or advise the user about possible memory leaks. Our invention covers dynamic recommendations.

    The Cork tool (http://www.cs.utexas.edu/~mjump/papers/cork-popl-2007.pdf) provides dynamic detection of memory leaks, based on heap analysis comparisons rather than garbage collection analysis. It requires JVM instrumentation to work and indicates only possible sources of memory leaks rather than suggesting GC tuning recommendations to users.
dynaTrace Diagnostics (http://www.dynatrace.com/en/Product.aspx) provides heap analysis but not GC tuning recommendations.

YourKit (http://www.yourkit.com/features/index.

GC tuning recommendations.

Summary:

    In summary, the prior art does not give GC tuning recommendations on a running JVM.

    The proposed approach here makes Garbage Collection tuning recomm...