MQL5 Cloud problems - very slow agents blocking genetic optimizations. How to solve it?

New comment
 
  • 17% (13)
  • 22% (17)
  • 10% (8)
  • 9% (7)
  • 10% (8)
  • 15% (12)
  • 17% (13)
Total voters: 46
 
Mql5 cloud is pretty awesome. One problem I have with it is how slow agents impact whole cloud, especially when using genetic optimization process. Usually 95-99% agents do their jobs very fast but you get stuck with 126/X or 127/X waiting for those last tasks to finish 25-100x longer than 90% of the agents to get to 128 tasks, so genetic algorithm can go to next step (and between each steps we get these lags from slow agents). It makes these optimizations very annoying, especially when testing longer period (few years every tick). How would you address this issue? Please leave your ideas how you think MQ could improve Cloud usage.
 

there are no ways to get around it unless MQ makes changes to how they send the task to the cloud.

I hope they can come out with a way that uses my COM more! 

 
First of all using the 64 bit architecture is a must in order to go further. Once this technical aspect is achieved, the others can be considered as well.
 

I get to use only agent from US, I would like to have access to more so as to speed things up.

 

I am not as concerned about the price for slow agents as this is so much faster than using my own computer, access to more agents would allow the whole job to complete quicker.

 

Is there some sort of diagnostic that would preclude machines with low speed or poor internet connections.? 

 

I wonder if the service desk or administrator of the Cloud would comment...... they may have some revisions already planned that will assist in the speed area.

 

For sure they want the tool to perform as that is one of the great features of MT5.

 

no speed with no agents.....?? 

 

This has been sitting like this for an hour.........!!!  What could cause that.?


now waiting 20 minutes for a connection.


It appears that the cloud service does not have the reliability for exchanging the information between computers and that may be a large part of the problem. whether the servers are inadequate or the software is not performing; I base that on the delays to connect.


It may be that the usage is greater than the capacity design.


I have sent some info to the service desk and they have yet to reply.. and I gather from that that they do not have the problems defined.


It has been very difficult to connect to the network for a couple days, clearly there is some problems that have to be dealt with by Metaquotes,  the network is not functioning properly.

 
Comments that do not relate to this topic, have been moved to "How to block login from different PC to the same account ?".
 

generally slow agents are not a problem when you do "slow complete algorithm" - you can just cancel when you get 90-95+ % of the results aborting the slow agents (you can run it again to get missing 5% much faster). The only problem is with the "genetic algorithm" - it is not fast as it should be and requires all 64/128 tasks to be finished to go to the next step of genetic algorithm.

On MT4 I was using genetic algorithm a lot. With MT5 MQL5 cloud I don't use it because of the speed issues - sometimes it gets completely stuck because of some one faulty agent.

 
I had a job running for ten hours and no connection to the cloud, last night...., the difficulty may not be with slow computers but with the network and the internet. One day, I am sure the people from Metaquotes will answer the questions and make a decision to improve the cloud function. I sent an email to the cloud folks and also sent a few log files to the service desk so that they could see some of the difficulties.
 
Blaiserboy:
I had a job running for ten hours and no connection to the cloud, last night...., the difficulty may not be with slow computers but with the network and the internet. One day, I am sure the people from Metaquotes will answer the questions and make a decision to improve the cloud function. I sent an email to the cloud folks and also sent a few log files to the service desk so that they could see some of the difficulties.

Sometimes the tester gets stuck while working with the cloud. You should try the following:

- forcefully close the metatester (ctrl+alt+del and close the process) 

- restart the terminal

- restart your computer

 

I have done each of those with limited success.

 

I feel that the system has to be 'debugged'.

 

I have sent numerous messages to the service desk. no replies yet.. which indicates to me that they have no solutions OR are not interested. 

12
New comment