Public discussion of the formula for calculating the cost of resources in the MQL5 Cloud Network - page 5
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
I didn't really mean trade. A CPU time buyer wants to buy himself N hours of 100 PR units. He has to somehow inform the server that I, Ivan_Ivanov want to buy N hours and am willing to pay M cents for it. Buyer puts an order in my personal cabinet to buy, if its price is higher or equal to some price, base, equilibrium for 120 days or whatever, then the buyer buys the processor time. The point is that the offers for purchase/sale of processor time put up through the site are both commands to the server to buy/sell and statistical data to determine the price. And the price chart is just that, for information.
It's too complicated - no one will actually make bids. The use should be simple - click on "start" and you're done. And the buyer should know that now the average price is X plus minus the delta. We will show the average price in the agent list window.
We have another problem - money entry (paypal, webmani and credit cards), where we will actually have to make manual transactions.
Renat:
We have another problem - money entry (paypal, webmani and credit cards), where you actually have to make manual transactions.
Manual transactions or manual control?
With cards and paypal I still understand, but what problems there might be with WM input (rather than withdrawal) is a mystery.
Or is "human" control of all financial transactions conceived?
0.01 cents per 100000PR per hour.
Why is that?
You can roughly put it like this. The number of registered participants at http://forum.mql4.com is something like 50,000. They all have a computer. This means that on average there are 2 cores per person, so 50'000*2=100'000 processor cores. Moreover, there are at least 10 times more MT users all over the world. And this is 100'000*10=1'000'000 cores. Besides, there is a category of MT users who have access to local networks from hundreds of computers, the percentage of these lucky ones is about 1% according to my calculations:
50'000*10*0.1*100*2+50'000*10*2=11'000'000 ядер.
To be willing to pay for reduced optimization time, you need a speed at least 100 times higher than for a single-core optimization. The number of users of MT that use optimization at each moment of time is no more than 1%, i.e. 50'000*10*0.1=50'000. The number of cloud cores per person:
11'000'000/50'000=220 ядер/чел. It turns out that the number of available cores is more than required by 220/100=2.2 times. -that's good. This is the peak load on the network, when everyone already knows about the cloud and actively uses it, at the beginning there will be much more available cores per person, I think 1000-10000 cores/person.
Now, one might ask, "How much would you pay for an extra 220 cores on your CPU to optimise 1 EA?" - how to find out?, you can ask this question in the member's profile in a special form, where max and min possible costs are marked (to adjust the limits over time), the price will be able to put up only those who have registered agents in the network (both sellers and buyers). Thus, you can display once a day the average cost per unit PR time (the figure will get very small, so better 1000PR per unit time).
That's how it is.
HH there is no point in calculating cloud computations based on iron depreciation costs, since the real computational cost of iron is 1000x lower (this includes 90% idle time, and as an example hordes of cloud scientific computing enthusiasts providing their computers resources "for nothing"). There is no way that 50'000 people can afford the depreciation costs of computers with 11'000'000 processor cores and rapidly aging hardware.
Which of the prices makes more sense from a joint compromise point of view of seller and buyer?
- 0.5 cents per hour
- 1.0 cents per hour
- 1.5 cents per hour
- 2.0 cents per hour
Please speak up.With cards and paypal I still understand, but what problems there might be with WM input (rather than withdrawal) is a mystery.
The problem is that users have to force themselves to register an account at MQL5.com and deposit money.
This is the very step that can decrease the number of users willing to use the service dozens of times. Psychology must be taken into account.
If the consumer is burdened with unknown parameters for the selection of prices, setting requests, etc., the number of users will be about zero.
The service should be very simple and reasonable. This is the only way it can attract buyers and sellers.
Imho, there shouldn't be any cheaper than the cost of PC power, 1 PC consumption is within 250W --> 1kW x 4 hours --> 6kW x day. The cost of electricity averages 8~10 cents kW
Yes, that's not a bad way to calculate the cost.
It turns out that the buyer will pay as much for the calculation in the cloud as he pays for electricity on his PC, but the calculation will be n-fold faster, where n-number of agents in the cloud.
And the vendors will be able to offset their electricity costs.
The result will be as follows: sellers will earn (savings on electricity, and the money saved is profit), and buyers will benefit in the speed of settlement (they will pay the same as they did before for electricity).
Another issue with free agents, which owners will run without being tied to logins.
Such free agents will be automatically and randomly allocated to jobs. But they will be available to customers who have a positive MQL5 account balance.
So, if you use a paid network, you get some free agents as a bonus.Renat:
If you burden the consumer with unknown parameters for selecting prices, setting bids, etc., then the number of users will be around zero.
The service should be very simple and intelligent. This is the only way to attract buyers and sellers.
Another issue with free agents, which the owners will launch without being tied to logins.
Yes, that's a very good way of calculating the cost.
That's weird, I thought I was going to say that again ;)
well then I'll expand on my previous post:
if 1 PC costs ~$0.6 per day in electricity consumption, you should allocate the cost ($0.6) of electricity correctly during the day, because from 8 to 17 hours user PCs are turned off by many, from 17 to 23 hours most turn on PCs, from 24 to 8 am PCs are turned off again. It turns out that the time from 0 to 17 hours should be the most paid, and from 17 to 24 hours a bit cheaper - I think the cost of selling resources during this time should match the cost of purchasing similar resources. Perfectly understand that even runet is distributed over many time zones, perhaps there will be no need to divide into time slots - there will be demand, there will be supply