Multi-core testing network for all comers - page 7

 
Renat:

If you allow everything to be counted on agents, the enterprising comrades will make a circus of mark-ups.

Definitely :)
 
gal:

Please explain why Metatester is going to the 137,138 UDP port of a remote client (NETBIOS) port description here:http://ports.my-addr.com/tcp_port-udp_port-application-and-description.php?port=137http://ports.my-addr.com/tcp_port-udp_port-application-and-description.php?port=138, when it has its own port 2000. From a security point of view, opening these ports is not a good idea.

You must have 137 138 open ports on your kernel.

I've been investigating this new service for a long time and haven't found anything like this!

only 1 port per core is open (you can choose it yourself)

--

 
Renat:

Only the actual machine time consumed is charged. If you have performed the tasks, the money is charged. If you don't, you don't have any tasks.

As tester-agents will connect to MetaTrader 5 Strategy Tester Poolers themselves, they will decide when they want to work and when they want to rest.


The cost will be calculated from the approximate formula = Time * Performance Rating * Price. When we get to the real valuations, we'll adjust the formula.

That's great!

If in the long run, I can set up a handler for the agent, that'd be great.

For example, I can plan because I know when my car is idle

and when I want to use my own processing power

and by configuring the pool's hours of operation with the sheller.

I can let the pool know when I'm putting in my resources...

 
Renat:
The cost will be calculated from the approximate formula = Time * Performance Rating * Price. When we get to the real estimates, we will pick up the formula.
A person can change agents' priority, can run other tasks in parallel after you have measured his PR rating, etc. There will always be some clever people who want to cheat the system.
 
It is precisely the pullers who will count the actual work, adjusting the cost based on average estimates. We are well aware of how and what they will try to inflate - methods of countering and controlling the results are laid down in advance.
 
YuraZ:

Wrong - you probably have some software running at the same time 137 138

I have been looking into this new service for quite some time and have not found anything like this!

Only 1 port per kernel is open (you can choose it yourself).

My software does not use 137,138 port, but if I open a port only in the metatest, I can not connect to me and recorded in the log firewall request to 137,138 port from a remote, and when I open 137,138 then successfully connected to my metatest, it probably should be checked on the machine that connects to me? But if the ports are used in parallel by some software, then why is the metatester blocked?
 
Put the metatester in firewall skip mode, it only works on one of its ports. And you are blocking the netbios protocol of the regular microsoft network. This protocol is used by the operating system itself.
 
YuraZ:

That's great!

If, in the long run, I can set up a handler for my agent, that would be great.

For example, I can plan because I know when my machine is idle

and when I want to use my own processing power

and I can set my own pooling hours with the shelfwalker.

I can keep the pool informed of when I'm giving up my resources...


I agree, a task scheduler for my agents wouldn't hurt.

SZZ and at the same time on the basis of data schedulers can make a rating of the pool loads.

 
Urain:

I agree, a task scheduler for my agents would be nice.

And at the same time, it will be possible to rank the pool's workload based on the schedulers' data.

The basic scheduler will be built directly into MetaTester - this will avoid external settings.

The MetaTester agent interface itself will change a lot - a lot of new features will appear.

 
When is the MQL5.community Pool scheduled to launch ?