During genetic optimization, only agents of one access point are used. It is connected with the specific features of the genetic algorithm.
The genetic optimization mode is
automatically enabled when the total number of optimization steps
exceeds one million in a 32-bit system and 100 million in a 64-bit
system.
Who can we reach out to at MetaQuotes to discuss how the Cloud Network works and how to improve performance of the cloud agents? I'm running over 200+ cores worth of agents and they don't stay active even when I triple or quadruple up on agent/physical core ratio.
bdclarke3: Who can we reach out to at MetaQuotes to discuss how the Cloud Network works and how to improve performance of the cloud agents? I'm running over 200+ cores worth of agents and they don't stay active even when I triple or quadruple up on agent/physical core ratio.
What do you mean by "don't stay active"? I think that MQL5 Cloud is working properly, only problem is that there is not enough backtesters (or jobs), to fully utilise whole network for long enough(for you to see).
And... increasing virtual agents are just stupid, you are paid for every invidual agent job, not whole lot togther. Ok, lets say, cloud will choosen more times you over my single quad core(only 4 agents), but that job will be so small, that you porblably will not get paid. And if there will be alot of jobs, your system will just work slower, per core and single agent will do less, becouse of all that "fake" agents and I bet ram also is quite limted per agent, since there are so many of them.
I would recomend only more agnets per core only if u get Hyper-threading cores, so you can fill them up, not more.
Anyways, my friend have plans to do some backtesting in near future, keep ya agnets ready otherways,I will lough, if my agnets will make my room warmer. :D
PS. I can't wait for next contest, I bet then will be a lot works in progress.
What do you mean by "don't stay active"? I think that MQL5 Cloud is working properly, only problem is that there is not enough backtesters (or jobs), to fully utilise whole network for long enough(for you to see).
And... increasing virtual agents are just stupid, you are paid for every invidual agent job, not whole lot togther. Ok, lets say, cloud will choosen more times you over my single quad core(only 4 agents), but that job will be so small, that you porblably will not get paid. And if there will be alot of jobs, your system will just work slower, per core and single agent will do less, becouse of all that "fake" agents and I bet ram also is quite limted per agent, since there are so many of them.
I would recomend only more agnets per core only if u get Hyper-threading cores, so you can fill them up, not more.
Anyways, my friend have plans to do some backtesting in near future, keep ya agnets ready otherways,I will lough, if my agnets will make my room warmer. :D
PS. I can't wait for next contest, I bet then will be a lot works in progress.
Even when the agents are in use by the CPU they don't even come close to peg'n it. Something tells me the program is limiting the amount of CPU usage. And yes I know the jobs are limited right now. I'm hoping this really picks up.
Contest? lol I have 300+ physical cores ready to go that do HT. Let's play.
2304
and,
Is it capable for multi client on one account?
I have a couple MT5s on a couple computers, can I use same account to pay for Agents from all of them simultaneously?
solved.
It used to be able to change agents while testing, but not any more,
that was the reason, maybe.
2 MT5s are using same account to use Cloud agents.
fantastic!
Admin
4535
and,
Is it capable for multi client on one account?
I have a couple MT5s on a couple computers, can I use same account to pay for Agents from all of them simultaneously?
Admin
4535
But now all of agents on Cloud Network says "authorized" and doing nothing...
I still have $105.98 in my account.
How can I make it "processing"?
2304
What do you mean?
Hi Rosh,
When I'm testing using Cloud Network,
for ex, MQL5 Cloud USA turn to "xxx agents of xxx used" and says "processing" on status column.
But most of time others like MQL5 Cloud Europe says "authorized" and "xxx agents available of xxx",
but not working.
Is there any way to force them switch to "processing"?
or,
How can I make all Europe, Europe2, HongKong, USA to work all together?
Admin
4535
Do you use genetic optimization? See help - http://www.metatrader5.com/en/terminal/help/tester/tester_agents:
2304
Do you use genetic optimization? See help - http://www.metatrader5.com/en/terminal/help/tester/tester_agents:
Indeed.
Mahalo!
Admin
4535
Please keep in mind, that your new tester agents will be shown in your Profile after they complete a task for MQL5 Cloud Netrwork.
171
403
Who can we reach out to at MetaQuotes to discuss how the Cloud Network works and how to improve performance of the cloud agents? I'm running over 200+ cores worth of agents and they don't stay active even when I triple or quadruple up on agent/physical core ratio.
What do you mean by "don't stay active"? I think that MQL5 Cloud is working properly, only problem is that there is not enough backtesters (or jobs), to fully utilise whole network for long enough(for you to see).
And... increasing virtual agents are just stupid, you are paid for every invidual agent job, not whole lot togther. Ok, lets say, cloud will choosen more times you over my single quad core(only 4 agents), but that job will be so small, that you porblably will not get paid. And if there will be alot of jobs, your system will just work slower, per core and single agent will do less, becouse of all that "fake" agents and I bet ram also is quite limted per agent, since there are so many of them.
I would recomend only more agnets per core only if u get Hyper-threading cores, so you can fill them up, not more.
Anyways, my friend have plans to do some backtesting in near future, keep ya agnets ready otherways,I will lough, if my agnets will make my room warmer. :D
PS. I can't wait for next contest, I bet then will be a lot works in progress.
171
What do you mean by "don't stay active"? I think that MQL5 Cloud is working properly, only problem is that there is not enough backtesters (or jobs), to fully utilise whole network for long enough(for you to see).
And... increasing virtual agents are just stupid, you are paid for every invidual agent job, not whole lot togther. Ok, lets say, cloud will choosen more times you over my single quad core(only 4 agents), but that job will be so small, that you porblably will not get paid. And if there will be alot of jobs, your system will just work slower, per core and single agent will do less, becouse of all that "fake" agents and I bet ram also is quite limted per agent, since there are so many of them.
I would recomend only more agnets per core only if u get Hyper-threading cores, so you can fill them up, not more.
Anyways, my friend have plans to do some backtesting in near future, keep ya agnets ready otherways,I will lough, if my agnets will make my room warmer. :D
PS. I can't wait for next contest, I bet then will be a lot works in progress.
Even when the agents are in use by the CPU they don't even come close to peg'n it. Something tells me the program is limiting the amount of CPU usage. And yes I know the jobs are limited right now. I'm hoping this really picks up.
Contest? lol I have 300+ physical cores ready to go that do HT. Let's play.