Discussion of article "Speed Up Calculations with the MQL5 Cloud Network" - page 6

 
366366:

I read the comments, some people write that in cloud network genetic optimisation will be slower than regular brute force optimisation.

Who is right?

Yes, it happens. But not always, and only if the number of passes is small.

And full brute force will be more expensive in any case.

 

I attempted to use the cloud yesterday and was connected to USA network...... and found that the speed was very, very slow. Actually, after a while it disconnected.

When it was connected the speed was far less than my own computer was processing both earlier and later in the day.


I have a question... was yesterday a bad day for connections or is there a current problem with the network.... The reason I am asking is that in the past I have had good results.


Thanks


daveM

 

I finally built my 24 cores system so cloud can get crunching.

CPU: 4X Opty 8439 SE
MB: Supermicro H8QME-2+
RAM: 16GB DDR2 ECC
GPU: Palit GTX580 3GB

Case: CM Stacker 810


 

Strategy tester uses 8 out of 24 agents. 

 This thing can't fit in single page - how cool is that? ^_^

 

 

Anyways so far I have seen 9 agents working, I can't wait for some massive optimisation job.

 

Update: Just got 16 in use. See attachment!

Files:
16ag.png  41 kb
 
CeleronS:

I finally built my 24 cores system so cloud can get crunching.

CPU: 4X Opty 8439 SE
MB: Supermicro H8QME-2+
RAM: 16GB DDR2 ECC
GPU: Palit GTX580 3GB

Case: CM Stacker 810

...

Anyways so far I have seen 9 agents working, I can't wait for some massive optimisation job.

 

Update: Just got 16 in use. See attachment!

Interesting. But it is a shame to use DDR2. And 16GB RAM isn't enough for such a beast
 

Well it's old hardware, but I must say DDR2 is underated, it doesn't keep me from playing any game what I want. I still can't believe that this system has 2x i7 4770 (hasswell) computing power. :O

As for a capacity for strategy tester, I saw 5.4GB max of all 24 agents combined. For gaming 16GB is still enough.  

 

 Anyways, hoping to see 100% utilization from cloud some day. :)_

Cores are fust enough to sustain 110PR. 

 
CeleronS:

Well it's old hardware, but I must say DDR2 is underated, it doesn't keep me from playing any game what I want. I still can't believe that this system has 2x i7 4770 (hasswell) computing power. :O

As for a capacity for strategy tester, I saw 5.4GB max of all 24 agents combined. For gaming 16GB is still enough.  

 

 Anyways, hoping to see 100% utilization from cloud some day. :)_

Cores are fust enough to sustain 110PR. 

Ah ok, I see. Let us know how it works.
 
Kindly, can I put MetaTester under a proxy ? Can I make some settings for network to use ? Is it possible to put my PC in the cloud and Run agents without network connection ?
 
Renat:

...

Renat, can you tell me if there are computers with 10 Gb of memory (or more) per agent ?
 
tol64:
Renat, can you tell me if there are computers with 10 Gb of memory (or more) per agent ?
One? Not likely. Look at the statistics by country, processor type and tasks performed for the last 24 hours...
Статистика - MQL5 Cloud Network
Статистика - MQL5 Cloud Network
  • cloud.mql5.com
Произведенный объем работы, распределение агентов по странам, RAM, CPU и битности OS
 
Rosh:
One? I don't think so. Look at the statistics by country, processor type, and tasks completed in the last 24 hours...
Thank you. Maybe it's time to build my own monster. ) Or maybe somewhere else offers services for volumetric computing....