About clustered parallel computation strategy testing support for mql4

To add comments, please log in or register
Levente Peres
91
Levente Peres  
Hi guys,

I was wondering, if any of you have ever used strategy testing / optimizing for experts written in mql4, using some sort of parallel computation grid solution?

I was thinking about using the free CPU time of my servers to aid me in strategy testing, and therefore I would build a computation grid, for example a Beowulf Cluster http://en.wikipedia.org/wiki/Beowulf_(computing) . Is there a strategy tester solution that can utilize that?

My expert is using a lot of CPU and so testing it for lengthy (1+ years) timeframe and many variations (50M+) takes forever even on a quad-core Xeon-based hardware. This is why I'm looking forward to build a cluster like that.

Levente
Ray
2646
Ray  

Wow.

Levente Peres
91
Levente Peres  
phy:

Wow.

OK... Really, I know how this sounded... :-) But still... Do you guys know of anything close to what I'm talking about? (Somehow I'm sure I'm not quite alone with this kind of problem, something like this could benefit everyone.)

I mentioned the Beowulf Cluster because that's the only solution I know and have implemented in the past in any serious way. I don't know a lot about MS based clusters though, for I'm mostly familiar with *NIXes. I even thought about trying to use the VMWare ESX Server infrastructure or similar... But I'm not sure if that would be the right framework for such a task...

Still, let's say we settle for one grid-computing solution or another, the question will still remain: is there a proper application, official or custom, that can take advantage of either infrastructural implementation?

Levente
Ray
2646
Ray  

I suspect what you really need to do is forget MQL4 for the moment, implement your strategy
in a suitable language that supports the performance options you desire, then transfer the optimizations
you discover back to an MT4 Expert for forward testing.

"You wrote on your Participant's page that you were using neural networks in your EA. How complicated is such an EA?
What is more difficult to create - an optimal neural network or the Expert Advisor's algorithm?

A neural network that forecasts the changes in exchange rates and a trading algorithm are inseparable, they belong each other.
Indeed, the Expert Advisor is very complicated. This trading system was first developed in С++, tested and optimized.
Only after that, it was re-written in MQL4 and debugged."

https://championship.mql5.com/2012/en/2007

Levente Peres
91
Levente Peres  
phy:

I suspect what you really need to do is forget MQL4 for the moment, implement your strategy
in a suitable language that supports the performance options you desire, then transfer the optimizations
you discover back to an MT4 Expert for forward testing.

"You wrote on your Participant's page that you were using neural networks in your EA. How complicated is such an EA?
What is more difficult to create - an optimal neural network or the Expert Advisor's algorithm?

A neural network that forecasts the changes in exchange rates and a trading algorithm are inseparable, they belong each other.
Indeed, the Expert Advisor is very complicated. This trading system was first developed in С++, tested and optimized.
Only after that, it was re-written in MQL4 and debugged."

https://championship.mql5.com/2012/en/2007

OK. This helped to adjust my perspective. Thanks! I'll to this. :-)
David
78
David  

i agree with you, shoul be fantastic, but shou be nice also if the metatrader team starts to develop a multicore versioneof MT4, at least for the optimization process

hunsheridan:
Hi guys,

I was wondering, if any of you have ever used strategy testing / optimizing for experts written in mql4, using some sort of parallel computation grid solution?

I was thinking about using the free CPU time of my servers to aid me in strategy testing, and therefore I would build a computation grid, for example a Beowulf Cluster http://en.wikipedia.org/wiki/Beowulf_(computing) . Is there a strategy tester solution that can utilize that?

My expert is using a lot of CPU and so testing it for lengthy (1+ years) timeframe and many variations (50M+) takes forever even on a quad-core Xeon-based hardware. This is why I'm looking forward to build a cluster like that.

Levente
To add comments, please log in or register