Speed improvements from ML Engine 075

Security has its price, the more security improvements you have that slower you will be generally. Some of our clients had problems with speed of security. We want to express here again that Security is the price paid for speed.

We have rewritten many parts of our backend to make ML faster with same security. We have implemented these improvements into ML Engine 1.2.075 which is released 20.Nov.2013.

The Proof

We are measuring speed by every release. Todo this we take a very large constant EA (which never changes) and process it with recent ML Engine and measure the speed. Todo this we have prepared special environment. This environment executes every produces Function in DLL 5000 times and measures speed for all exported functions inside DLL

Since EA is always same, we do have same amount of functions by every update, the only difference is ML Engine changes (the backstage work of those functions).

Here is a test on 19.Nov.2013 made between ML Engine 074 (which is current) and ML Engine 075 (release on 20.Nov) and the results:



We let run 5000 iterations on same broker, same mt4, same amoung of functions with two different ml engines and as you see results are very good. On Phase5 we have runned 1865 functions * 5000 iterations = 9.325.000 times and it took our machine 22 seconds to finish with 074 and 11 seconds with 075.

Conclusion

We will keep our improvements and we will try to optimise our backend more in future to ensure better speed with even better security measurements.