MQL5 Cloud Network (2)

Borrow CPU power from the cloud

In the previous article, we were mainly the ones who lent CPU power and profited from it, but this time we were the ones who borrowed it and tried to see how much faster the test could be. It is extremely fast! The own machine + local network alone uses 11 CPU minutes, but it is so fast that it is not even comparable to that. The above image is the result of about 10 minutes of execution, but in terms of the number of completed tasks,

The difference was about 18 times.
*The variation in the number of executions within CPUs in the local network is assumed to be due to resources being temporarily provided to the MQL5 Cloud Network.

How much money will it cost?

I used it for about half a day and it cost about $5.
(The optimization process is completed fairly quickly, so I changed the time frame and period several times while running the optimization.
Until a decade ago, there was some resistance to spending money on valuable tools such as AWS and paid software to increase productivity, but as I got older, I began to feel the weight of money < < < the weight of time, and the ability to shorten time with a small cost was very appealing.
The most valuable thing I was looking for in MT5 was that it exceeded my expectations.
I usually have a lot of time to play with my PC while I am busy with my day job, so when I think that I can speed up my backtesting with the profit I get from lending out my resources during that time, I feel like I am saving time itself.

Method of payment of expenses

Although the benefit was automatically offset from the profit made by providing CPU power to the MQL5 Cloud Network in advance, the following procedure is required for first-time users.

MQL5 Cloud Network, steps to buy resources

• Register with MQL5.community ;
• Deposit through the payment systemor get credit through jobs, marketsor signal services etc ;.
• Set up a login and password in the terminal settings.

What is the payment rate?

This was also mentioned in the MQL5 FAQ.

What formula is used to calculate agent compensation? The rental rates for individual agents are based on their PR and the time spent on the tasks ordered by the network user. The remuneration for a task completed by an agent is calculated comprehensively using price, productivity, and time as the factors. Money=Price*PR*Time, where: Money is the reward for the completed task, Price is the PR cost per second, which is the same for all agents in the MQL5 cloud network. PR is the agent's performance rating, calculated in a special way. The higher an agent's PR, the faster it completes tasks and, consequently, the higher its per-agent rental cost. Time is the time taken to process a task; the higher the PR value, the faster the task is completed. The calculated cost of a task does not depend on the PR value of the agent processing the work ordered by the network user. This is because compensation is determined by both the processing time and the PR value of the completed task. What is the cost of 1 PR per unit of time? The cost of an agent with PR=100 is 0.01 USD per hour. 1 quant is a unit of work. 1 quant corresponds to 1 ms (1 millisecond) of work for an agent with PR=1. Thus, the cost per quant can be calculated as follows Quant price = 0.01 USD/(100PR*3 600 000 ms)=2.77778E-11 [USD/(PR*ms)] The table below shows an estimate for one single-core tester agent (with PR=100) running for one hour and one month.

Period	Quant Price, USD/(PR*ms)	Agent, PR	Time, ms	Total, USD
1 hour	2.77778E-11 100	100	3 600 000	0.01
1 month	2.77778E-11 100	100	2 592 000 000	7.20

important point

When using the MQL5 Cloud Network, programs that use DLLs inside the EA will not run on an external PC.

Indicators and EX5 libraries enabled using directives such as #property tester_indicator and #property tester_library (DLLs are not allowed in the cloud)

https://www.mql5.com/ja/articles/341 When I actually try to run the program, it seems that although a request is made to a PC on the network, the process fails immediately and terminates.
When I check the log file of the PC on which the request is executed, the following message is displayed.

IK   0   18:09:27.969    Tester  expert file added: Experts\xxxxx.ex5. 219060 bytes loaded
DR  0   18:09:27.984    Tester  file added: Libraries\yyyyy.dll. 673814 bytes loaded
OQ  0   18:09:27.984    Network 17976 bytes of optimized inputs info loaded
OD  0   18:09:27.987    Tester  successfully initialized
QF  0   18:09:27.987    Network 579 Kb of total initialization data received
PJ  0   18:09:27.987    Tester  Intel Core i7-4770K  @ 3.50GHz, 7872 MB
FO  0   18:09:28.233    Tester  optimization pass 299062603 started (batch of 2 tasks)
JF  0   18:09:28.290    Symbols EURUSD: symbol to be synchronized
LS  0   18:09:28.348    Symbols EURUSD: symbol synchronized, 3624 bytes of symbol info received
GK  0   18:09:29.087    History EURUSD: load 27 bytes of history data to synchronize in 0:00:00.021
HE  0   18:09:29.087    History EURUSD: history synchronized from 2005.01.03 to 2019.04.04
DR  0   18:09:29.242    History EURUSD,M1: history cache allocated for 4191148 bars and contains 343091 bars from 2008.01.02 09:01 to 2008.12.31 20:00
PM  0   18:09:29.242    History EURUSD,M1: history begins from 2008.01.02 09:01
NP  2   18:09:29.245    Tester  2009.01.01 00:00:00   DLL loading is not allowed
ND  2   18:09:29.245    Tester  global initialization failed
QL  2   18:09:29.245    Tester  global initialization critical error
OI  3   18:09:29.404    Tester  working time at stop is 0:00:00.298

For complicated processes or processes that are difficult to implement with MQL5 alone, we would like to use DLLs, but this is a bit of a problem that we have to deal with.
However, it is quite risky to have DLLs sent from other people's PCs without their permission and to be allowed to execute them, so this level of inconvenience may be necessary.

summary

The processing speed is much faster than MT4's single threading, and the small cost buys time, which is a very welcome feature.
I will actively use this feature in the future!