Hi, everybody. I regularly notice various problems with MQL5 Cloud, namely:
- Super-long wait for the start of testing, despite the fact that the selection of agents by the system and connection to them takes place in a matter of seconds. Moreover, everything is displayed as if everything is in order and the tests are going normally, but you can wait for the real start of testing in this state for more than one hour. At the same time, the system does not forget to write off money on the sly.
- Testing is extremely slow and it is not affected by the number of operations being tested, the amount of Expert Advisor code, the period or type of optimization, or any other factors that I could influence. The screenshot below shows the optimization time that the tester calculated in the genetic optimization mode when testing for a period of two weeks (!). In a normal situation, the optimization of the cloud would take less than a minute.
Of course, this does not always happen, often the cloud works out perfectly and the optimization process is very fast and cool, but I meet such problems so often that I decided to ask here. Share your experience, whether someone is faced with such a thing, how did you solve this issue for yourself? Can it be cheaper and faster to rent a server and perform calculations there?
ververy
Renat Fatkhullin:
In the last two weeks, we have been doing a lot of work to optimize the cloud network and fix long-standing errors.
A lot of things have been fixed both on the part of servers and agents with terminals.
Be sure to use the latest terminal betas(2515).
With the release of the release, we will restart the distributed computing network and show its advantages.
At the moment, the cloud network has no alternative. Without it, it is almost impossible to conduct research.
Thanks for the news, sounds promising.
wenay
Renat Fatkhullin:
In the last two weeks, we have been doing a lot of work to optimize the cloud network and fix long-standing errors.
A lot of things have been fixed both on the part of servers and agents with terminals.
Be sure to use the latest terminal betas(2515).
With the release of the release, we will restart the distributed computing network and show its advantages.
At the moment, the cloud network has no alternative. Without it, it is almost impossible to conduct research.
zhdems
renat
In the last two weeks, we have been doing a lot of work to optimize the cloud network and fix long-standing errors.
A lot of things have been fixed both on the part of servers and agents with terminals.
Be sure to use the latest terminal betas(2515).
With the release of the release, we will restart the distributed computing network and show its advantages.
At the moment, the cloud network has no alternative. Without it, it is almost impossible to conduct research.
ververy
After testing previous versions of my Expert Advisor, I concluded that the problem is really somewhere in it, not in the cloud. As soon as I find the reason, I will write. I hope it helps someone
antt
You have gentic optimization. The calculation time of one generation cannot be less than the calculation time of the slowest pass of this generation. Your Expert Advisor can actually be tested for half an hour, with a certain combination of its parameters.
ververy
The Expert Advisor is compiled, of course. This has nothing to do with it.
barabashkakvn
The main rule: before sending a task to the cloud, you need to compile the Expert Advisor. That is, we compiled -> ran the test in the cloud. There should be no debugging runs between these two steps.