MetaTester5 Agents Manager Seems Seriously Broken

 

Hello all, first topic/post in here. 

Joining the forums today to ask some questions about the Tester Agent manager for the MQL5 Cloud Network.

I've had the program running nonstop for the past week or so to test the amount of network traffic and see how well it's being utilized but it seems as though there are some serious issues with how it optimizes its use on systems. In addition, there seems to be next to no settings to customize what resources you want to have at an agents disposal and what the limitations are on each.

As an example: I sat and watched TWO Tester Agents take off and utilize 30GB of RAM each for a solid 4 minutes effectively hamstringing the entire system (this basically removed the possibility for any of the other 30 agents to use RAM resources), and after all computation was done the MQL5 Cloud Network "Time Taken" hadn't increased a single minute. Now, I understand that utilizing the tester agents isn't exactly something for people to rely on for income, but to see how there is no way to set any of these things and pre-define what agents can use seems a bit silly. Not only that, but the fact that agents are currently capable of utilizing ALL of a systems RAM with a minuscule return (Less than .01 cent per minute) and not properly registering the usage time on the network.

There has been far more computational time registered on the computer than what has registered on the MQL5 Cloud Network, and I'm assuming it's a collective sum of all agents time computing on the network. (all 32 agents on this system added up in overall compute time)

Side Note: As I'm typing this, I just witnessed another round of computation utilizing 16+ agents each running for about 30 seconds each and 1 minute of "Time Taken" was added. Just doesn't seem right. Maybe I'm not seeing something here. Lets chat!

 

There is one good summary thread:
All (not yet) about Strategy Tester, Optimization and Cloud

This is small thread with all the information collected.

 

Hi Sergey!

Nothing in there describes why Agents are able to take nearly ALL of the computer resources rendering the other Agents useless, and at the same time not paying out more for the ridiculous amount of resources they are utilizing. 

When 2 Agents take 100% of the RAM the other 30 agents on this computer are crippled and I have had them disconnect because the system was effectively crushed. On top of that, the logging of computation time is not calculating properly.

 
There is some news from MetaQuotes (from rus part of the forum) - I think it may be related to this subject -
Look at #1 
or - 

Forum on trading, automated trading systems and testing trading strategies

1957. Tester rips off all memory

Renat Fatkhullin , 2018.12.06 11:33

We found a memory leak and we will release an update on MetaQuotes-Demo in a couple of hours.

The cloud network was temporarily stopped and will be updated first.

... 

 

Forum on trading, automated trading systems and testing trading strategies

1957. Tester rips off all memory

Renat Fatkhullin , 2018.12.06 14:29

We released version 1958 with the correction of unpack error and memory leak in MQL5:

  • available on MetaQuotes-Demo and MetaQuotes-Beta
  • will be available in a short time on Help -> Check updates -> Beta
  • rollout agent updates in the cloud network


Reason: