MetaTrader 5 Strategy Tester and MQL5 Cloud Network - page 22

 
Interesting:
Better not, although if you put a few do not do more than two.

Hmmm not clear, I have agents with a small PR (with divided cores) - they scale by the number of tasks completed

CPUDescriptionIPPRTasksRevenueCreatedActivity
Intel Celeron 2.40GHz, 447MB

3032502011.07.192011.08.27
Intel Atom D510 @ 1.66GHz, 2031MB

414702011.08.262011.08.27
Intel Atom D510 @ 1.66GHz, 2031MB

4112702011.08.262011.08.27
Intel Atom D510 @ 1.66GHz, 2031MB

87902011.08.262011.08.27
 
IvanIvanov:
The logic is clear, the main thing is understanding, a common algorithm would be like this - bought - tested in cloud mode - ran out of money - finish on your capacity, but without bummer at the end of the test......

Also, the terminal has automatic cache of results, which allows to stop testing in any place, and then when restarting (if conditions, periods and settings have not changed) - continue calculations from the last place.

Note lines at the end of calculations in the log "from file cache used XXXX results".

Try to check it yourself.

 
IvanIvanov:

Will there be a charge for text messages?

System SMS messages are always free of charge:

  • confirmation of milestones
  • confirmation codes for financial transactions
  • balance enquiries
 
IvanIvanov:

Hmmm not clear, I have agents with a small PR (with divided cores) - scale by the number of tasks completed

CPUDescriptionIPPRTasksRevenueCreatedActivity
Intel Celeron 2.40GHz, 447MB

3032502011.07.192011.08.27
Intel Atom D510 @ 1.66GHz, 2031MB

414702011.08.262011.08.27
Intel Atom D510 @ 1.66GHz, 2031MB

4112702011.08.262011.08.27
Intel Atom D510 @ 1.66GHz, 2031MB

87902011.08.262011.08.27

The point is that under a good load all agents, of which there will be more cores, will be engaged. This will cause the CPU to run at 100% of load all the time (theoretically, with constant connectivity - 24 hours a day * 7 days a week).
 
IvanIvanov:

Hmmm not clear, I have agents with small PR (with divided kernels) - scale by number of executed tasks

It's all within the margin of error + some of the agents got a very slow EA in the calculations, which spent more time, resulting in executing fewer passes in the same time.

On longer intervals and masses of different tasks, the statistics level off - here's a piece from my agent table:


 
Renat:

This is all within the margin of error + some of the agents got a very slow EA in the calculations, which spent more time, resulting in executing fewer passes in the same time.

On longer intervals and masses of varied tasks the stats level off - here's a piece from my agent table:


I've got guys drooling looking at your OS... :-)

Is this the server?

 
Renat:

The terminal also has an automatic result cache that allows you to stop testing anywhere and then when you start again (if the conditions, periods and settings have not changed) continue calculations from the last place.

Note the lines at the end of the calculations in the logs "XXXX results were used from the file cache".

Try to check it yourself.

This needs to be spelled out in the submission conditions! From what you wrote in the first sentence I understand 80%

From personal experience, Wouldn't it be easier to redirect to completion on your own resources? It would encourage you to have a reserve on your balance sheet.

 
IvanIvanov:

I've got guys drooling over your operating system... :-)

Is it a server?

It's basically a server config, but it's one of my personal work computers. 24 cores, 24gb of memory + a bunch of stuff.
 
IvanIvanov:

In my personal experience, it's easier to redirect to completion on your own resources, it will encourage you to have a reserve on your balance sheet.

Everything long ago and originally worked automatically in agent hot-in/out mode at any time (even in the middle of a test).

Try running a test with claud, then disable claud ticking - see that the tests will continue on local agents. If the claud is disabled for lack of money, the calculations will continue on available agents (local or remote).

 
Renat:
Essentially a server config, but this is one of my personal work comps. 24 cores, 24gb of memory + a bunch of stuff.

:-) Showed your reply to mine - asked when I would have one, ... smiled..... she works in my shop with computers....

Are these 24 in the same case or not?