Agents consuming Ram

 

I had to uninstall the agents on my computer recently as they began using all of the RAM my computer had to offer. There was no clear cause for this, and they used so much of my computers RAM that it caused programs up to and including my antivirus program to crash. It made my computer unusable in a way that it had never done before. With 64GB of ram, no one program should be ever be using all of it. I would love to know if there is a fix for this so that I can go back to making a bit extra selling my processing power here.


To be clear, I did attempt changing the schedule to at least stop this happening while I'm at my computer, but this did not resolve the issue at the time. Until the problem is resolved, I will certainly not be putting the agents back.

 

I've had the same problem since I started selling computing resources last year. On top of that, agents will sometimes fill up the hard drive they're installed on. I expected agents to max out CPU, but not RAM and storage. Maxing those two leads to system instability, and I see no way of limiting them.

Service Desk doesn't offer help with agents, instead redirecting us to the forum, where apparently this has gone unanswered for years.

Please devs, teach agents some boundaries or let us set them.

 
androgendo #:

I've had the same problem since I started selling computing resources last year. On top of that, agents will sometimes fill up the hard drive they're installed on. I expected agents to max out CPU, but not RAM and storage. Maxing those two leads to system instability, and I see no way of limiting them.

Service Desk doesn't offer help with agents, instead redirecting us to the forum, where apparently this has gone unanswered for years.

Please devs, teach agents some boundaries or let us set them.

The problem is that the workloads run on the MQL5 network are often designed for servers, where a server will (nowadays) have 1-4 TERAbytes of ram equating to 20-100 GB of ram per core (or agent). Remember folks the primary difference between a server (or heavy duty workstation) and a regular computer is the memory storage and other I/O.

When a "regular" computer encounters this it is hopelessly outmatched in the storage department by the massive arrays of information often containing millions or even billions of data points that HAVE to be stored to run, even if its processor is sufficient to handle the actual calculations.

There is nothing MQL5 can do to prevent this short of putting direct constraints on the programs that their computing platform is allowed to run.