Memory profiler fails on large memory snapshot

[rnbguru]

I have a wcf service I am trying to watch through ants profiler 7. The service grows rapidly, as I am trying to identify a memory leak. Within two minutes, it grows to two gig of memory. When I then click take a snapshot in ants profiler, the profiler spins with "taking a snapshot this may take a few minutes" for upward of an hour. Unfortunately, I had to kill the process at the hour mark as it was hurting a production level box.

Looking at the machine's cpu/memory usage during the hour of the snapshot, the cpu went from running at 80% while profiling to a flatline of 10% while it was taking the snapshot. What's more, the memory dropped the full two gigs 20 minutes into the snapshot. These two things combined make me believe the profiler stopped taking the snapshot and simply spun after that point.

Do you have any idea what might be causing this?

[Brian Donahue]

All I can suggest is there is not enough memory to deal with all of the results data.
_________________
Brian Donahue
Technical Support
Red Gate Software Ltd.
44 (0)870 160 0037 ext 8521
US and CAN 1-866-RED GATE ext 8521

[rnbguru]

Do you know how much memory loading the results takes? Is it roughly one to one with the size of the process memory?

Also, would there be a way to then instead, simply dump the screenshot to file bypassing memory, so I could possibly load it on another box where memory might be more readily available to render it all?

[Brian Donahue]

Apparently it has to do with the number of objects being tracked more than anything -- about 50 million seems to get things going slowly. Results woul obviously be better on x64 (profiling a 64-bit app as well, I suppose).
_________________
Brian Donahue
Technical Support
Red Gate Software Ltd.
44 (0)870 160 0037 ext 8521
US and CAN 1-866-RED GATE ext 8521