Crazy cache of testing agents - page 9

 
Renat Fatkhullin:

The difference in economic and technical levels of knowledge is huge and so you cannot pass them on.

Unfortunately.

Especially if one has to make others believe this.

Here https://cloud.mql5.com/ru/faq/payments YOU indicate that you are willing to listen to reasoned opinions on the price of the service:

The price is set too high/low and needs to be changed

The question of a fair price, which satisfies both buyers and sellers of computing resources is still under consideration. We are looking for a balance between the interests of the customers and the owners of the testing agents. The main focus is on the fact that idle CPU capacity can generate a small but steady income.

Without participation in the MQL5 Cloud Network, the computer already consumes resources, but 99% of the time it does not produce anything. So make it be somewhat useful!

And in practice it turns out that you are only willing to listen to yourself and you get sidetracked from discussing the merits of the issue.

Earlier, I made a calculation and justified my assumption about the development of the resource in the case of lower prices, but instead of expressing your position, you literally hint "You are all illiterate fools, while I am Star and will shine to whomever I want" - surprised and saddened by this position. I admit that I don't know something, while you, as stated earlier, deny the very possibility of being wrong.

Regarding the cache, and the plausible performance degradation when reading a single piece of data - there is memory caching software, for example this FancyCache , which, in theory, solves the problem.

 
One thing I've learned in all this time is that arguing with developers, and especially asking them for something, is a completely futile exercise. Sometimes you can run into unflattering "eat what you're given" kind of pitches. Sorry, I'm fed up.
 

Decided to try the cloud today - did some optimisation and here's the result

I am shocked how expensive it is! In fact there were 7128 passes - because some of them had wrong input parameters and returned with such an error

2017.11.25 04:53:41.798 MQL5 Cloud Europe 1     pass 9672 tested with error "incorrect input parameters" in 0:00:03.688 (PR 95)
2017.11.25 04:53:42.528 MQL5 Cloud Europe 1     pass 9724 tested with error "incorrect input parameters" in 0:00:01.328 (PR 94)
2017.11.25 04:53:42.710 MQL5 Cloud Europe 1     pass 9869 tested with error "incorrect input parameters" in 0:00:00.485 (PR 131)

If you even take the calculations of MQL, and divide the time of agents by 6 cores (on average let it be so), then it turns out that for 13 hours of average computer I paid 7 dollars and 53 cents? 56 cents an hour? 4,967.91 for the year? For that kind of money you could buy 10 buu computers to optimise!

The service is way too expensive, it's not reasonable to use it! You have to reduce the cost at least five times, and in general 10 times - because the amortization of PCs according to the standards is 3 years, while in reality the average life span of a PC is 6 years.

What about mining on video cards - here is the Klondike! Only agents are idle, and the profit is on paper...

 
Aleksey Vyazmikin:

Decided to try the cloud today - did some optimisation and here are the results

Where can I see these statistics, I couldn't find them myself?




 
Vitaly Muzichenko:

Where can I see these stats, couldn't find them myself?



Profile -> Agents -> Tasks

 
Aleksey Vyazmikin:

Profile -> Agents -> Tasks

I'll just get to the end of the test now and post the results


Here are my results


 
Vitaly Muzichenko:

I'm just about to get to the end of the test and I'll post the results


Here are my results



How is it that your average cost per hour for an average computer is significantly different from mine, and different from each other!

On the first three passes #1154 - 3.7; #1068 - 3.29; #525 - 1.13.

Pity they didn't give the figure you have on the first screen or is it genetics there?

In general, the price spread and you have a big, even if you do not look at the time, but the number of passes

For one pass:

0,002279
0,001781
0,003485


It seems that we are being misled with pricing!

 
Aleksey Vyazmikin:

How is it that your average cost per hour of an average computer is significantly different from mine and different from each other!

On the first three passes #1154 - 3.7; #1068 - 3.29; #525 - 1.13.

Pity they didn't give the figure you have on the first screen or is it genetics there?

In general, the price spread and you have a big, even if you do not look at the time, but the number of passes

For one pass:

0,002279
0,001781
0,003485


It seems that we are being misled with pricing!

Tests in genetic algorithm mode. I enabled testing without the cloud, it took a very long time, but the test itself only did 5%. That is, if you don't use cloud and buy personal machines for this purpose, and get not new but CU ones, then you need super a lot of optimizations to make it all pay off, and most likely it won't pay off very soon. What about pricing, I don't know, but I like the cloud.

 
Vitaly Muzichenko:

Tests in genetic algorithm mode. I switched on testing without the cloud, it took a very long time, but the test itself only made 5%. That is, if you don't use the cloud and buy personal machines for this purpose, and get not new but used ones, you need a lot of optimizations to make it all pay off, and most likely it will not pay off very soon. What about pricing, I don't know, but I like the cloud.


I run my computers (4pc) about 280 days a year, both on MT4 and MT5. And it's the more profitable option, judging by the rates... About 10,000 passes will take me 8 hours, and I will save 7 quid :) I usually have something to occupy my time, such long optimizations are easy to put on overnight.

And most importantly, I'd like to get comments on price spread from @Renat Fatkhullin.


 

To clarify, 10560 passes took me 8 hours 56 minutes 52 seconds, which means I could have spent $11 15 cents in the cludes. That's something very expensive for me...