Profiler Extension - Timelines of Memory Snapshots

[mikejwatts]

Briefly: the profiler is a good tool but I take snapshots and analyse them in a spreadsheet. It would be a good idea to automate this. It could be done by reading the files produced and after requesting and saving a snapshot. Maybe there is an API available to request and retrive data? Or should I make extensions to the profiler [code]. A pointer in the right direction would be appreciated.

Mike Watts

[Petr Cyhelsky]

Hi,

mikejwatts wrote:

[mikejwatts]

Petr,
Thanks for replying. I am looking at an application which runs constantly monitoring call quality from all over the UK, data coming in over serial lines i.e. from monitoring hardware, not input from human operators. What you suggested may help but the big problem is the amount of data and I would presumably need to manually press the button to convert data to CSV.

The generation and collection of objects is dependent upon what data is coming in ( and other factors ) but if I could make a graph of a particular class's instance count over a period of time I should see a steep rise at the application start and some cyclic variation after that.

For a start I can discount any classes that have instance counts that do not vary during a particular testing scenario. After that, ( for a cyclic variation ) if the peaks and troughs stay at the same-ish level over a reasonable period then they probably indicate there is no leaked memory.

Any class instance counts that don't follow those two patterns need to be considered further.

If I could analyse the data in a Java program then I could provide this level of data reduction and do some graphing etc. The version I have doesn't ( appear to ) have either of the features you mentioned so I'll try those first but I think I'll need to do a bit more automation before I can access the figures I want.

Cheers, Mike Watts