Backtest/Optimization freeze-up

[Dionysus]

The dreaded issue from nt7 has once again reared its ugly head, but I'm hardly surprised, as I'm not even certain how avoidable such a thing is. .

This is a very broad level problem, so I'm not sure what details or documentation I can provide, but I've finally found the time and have also created the necessary resources to run some rather robust and complex backtests and optimizations on nt8. The speed is certainly improved, however, when the tests get too complex (too many parameter combinations, too many instruments being tested, too many bars being tested such as 1min increments back to 2007, etc), I'm once again encountering the freezing-up of ninja. . and the 'abort' button doesn't work, because the program is just *that* frozen. I tried waiting about 20 minutes, and clicking abort the moment it un-froze for a few seconds at a time, but had no luck with this, and had to eventually use task manager to force a shut down.

Completely preventing any temporary freezing behaviors from incredibly complex or data intensive tasks may be nearly impossible, but if it could at least allow for a proper 'abort' in these cases, this would be an immense step forward. . .

[NinjaTrader_Matthew]

Our developers would like to see your scenario where you're hitting a total lockup. Can you send in your strategy (or a rough sample strategy) with the settings you've used?

[Dionysus]

Sorry for the late reply. . its been slow because I'd hoped to duplicate the conditions for you guys, as I can't send over my strat/indicators unfortunately, as they are proprietary and quite valuable. . . my apologies. I still hope to find the time to emulate the situational conditions so that I CAN provide you the needed materials to reproduce the error, soon

[NinjaTrader_Matthew]

Ok thanks for the update - we're unaware of a scenario where we would hit a total lock up so if you can reproduce that, definitely let us know.

Regarding the abort operation, as long as the optimization is not locked up, it should truly abort the run. The issue would only occur if one of the iteration was stuck in some sort of loop due to the settings being optimized.

[gregid]

I can press the Abort button just fine but it usually takes about 10 min for the optimization to truly abort.
Then it doesn't free up the memory used by Strategy Analyzer and in fact it increases memory consumption, suggesting memory leakage (even after closing Analyzer window) quickly reaching 100% memory consumption. Every time I run optimization I need to kill NT process as my machine is becoming unusable.

[NinjaTrader_PatrickH]

Hello gregid,

Thank you for your post.

The .NET garbage collector would need to be triggered and there is no true and proper means to do so without restarting the program itself.

[gregid]

Thanks Patrick,

Can you help me and clarify if I understood you correctly:
Does it mean the memory leak is simply expected at every run of the optimization?

[NinjaTrader_PatrickH]

gregid,

A memory is not expected at all. However, an example would be running a backtest, editing your code, then running another backtest. This would result in the memory holding the previous version of the strategy.

You may wish to try running with IsInstantiatedOnEachOptimizationIteration: http://ninjatrader.com/support/helpG...niteration.htm

[gregid]

Thanks Patrick, you got me scared for a moment