Distributed Hardware Evolution Project

Post Reply
User avatar
bcavnaugh
Posts: 77
Joined: Fri Sep 29, 2017 12:06 am

Distributed Hardware Evolution Project

Post by bcavnaugh » Tue Oct 23, 2018 5:14 am

Distributed Hardware Evolution Project Might be a good Project to add for 2019
https://www.dhep.ga/boinc/
https://www.dhep.ga/boinc/server_status.php
Image
Crunching@EVGA The Number One Team in the BOINC Community. Folding@EVGA The Number One Team in the Folding@Home Community.

marsinph
Posts: 88
Joined: Wed Apr 25, 2018 6:09 am

Re: Distributed Hardware Evolution Project

Post by marsinph » Tue Oct 30, 2018 8:55 pm

Hello,
I don't think so !
Why ?
Not about the project itself. I also not understand the SCIENTIFIC goal of this project !!!....
To build a virtual "island" and manage citizens...!?!?!?!?

But about credits and very very long running time (weeks on a intel i7-2600 at 4.00Ghz !!!)
On personal stats, the credits are given, each hours, step by step. It need to be manualy done,
because no synchro with BAM. The project says it very clear. They not follow
the rules. No problems, we know it.

But only on project it self and on local account manager.

Then once again, WU need some weeks to completed. So, in each case, can not participate at Sprint.
Then demay for terurning WU is six months !!!
Yes six months !
And there is no checkpoint ! So WU restart and restard again.
Not the credits. They stays, but there are only individual.

So?, for all those reasons, I think it is not a nice idea.
If I good remember, a few years ago, a project was removed, because it was not scientist. With reasons.

Best regards

Woodles
Posts: 20
Joined: Wed Dec 21, 2016 11:15 am

Re: Distributed Hardware Evolution Project

Post by Woodles » Wed Oct 31, 2018 3:34 pm

No offence but just because you don't understand what a project is doing doesn't mean that it's not doing any scientific work.

The goal is machine learning. Each "island" with it's "citizens" is a specific problem set up with a number of parameters and an algorithm ran to see if it can reach a stable solution. It's evaluating different algorithms and their possible usage.

The tasks downloaded to the host are placeholders, all work done is returned to the project every fifteen minutes by a trickle up process and credit awarded every hour, the work units don't need to be finished.

I don't use BAM but if there is no synchronisation then that sounds like BAMs problem, stats sites follow the project without any problems and BOINC manages fine on a local machine.

Again, the WUs are effectively fifteen minutes long not weeks. Think of it as finishing after fifteen minutes and immediately downloading again to continue the work.

Checkpointing is done on the project server ... every fifteen minutes!

It's just as suitable as any other project for Formula Boinc inclusion although maybe not for a sprint. I don't know if their server could cope with thousands of hosts connecting every fifteen minutes!

Bryan
Steering comitee
Posts: 14
Joined: Thu Dec 04, 2014 3:47 am

Re: Distributed Hardware Evolution Project

Post by Bryan » Wed Oct 31, 2018 3:53 pm

@marsingh - you don't understand how the WU are actually working. Yes, the WU will run for months but they can be stopped, restarted, or aborted at anytime without penalty. The WU are just "placeholders" so BOINC knows how to handle them. It is irrelevant how long they say they are going to run. Whenever you are running a WU, the project sends information to the computing executable and then as work is being done it is reported back to the project periodically. That is the reason you get credits even if you haven't "finished" a WU. These aren't WU aren't like most projects where there is a beginning and an end. They are "dynamic" WU meaning they change the values they are computing.

This is the similar to the way that Climate Prediction works. They have very long WU that will actually finish at some point - sometimes months. However, while you are running a WU your machine sends information back to the project periodically and you get credits for the work completed to that point. These reports going back to the project are called "trickles".

Skillz
Posts: 53
Joined: Wed Sep 27, 2017 11:51 pm

Re: Distributed Hardware Evolution Project

Post by Skillz » Wed Oct 31, 2018 9:26 pm

IIRC, the "deadlines" are only there because BOINC requires it. Otherwise, they wouldn't even have deadlines.

marsinph
Posts: 88
Joined: Wed Apr 25, 2018 6:09 am

Re: Distributed Hardware Evolution Project

Post by marsinph » Sun Nov 04, 2018 5:55 pm

Hello Bryan,
Thank you for explanation.
It were afterwards given by project developer.
I know and see about credits. Depending if I crunch or not, average increase or decrease.
Credit stay increasing. No any problem.

Then Bryan, please consult private message.
Thank you



Bryan wrote:
Wed Oct 31, 2018 3:53 pm
@marsingh - you don't understand how the WU are actually working. Yes, the WU will run for months but they can be stopped, restarted, or aborted at anytime without penalty. The WU are just "placeholders" so BOINC knows how to handle them. It is irrelevant how long they say they are going to run. Whenever you are running a WU, the project sends information to the computing executable and then as work is being done it is reported back to the project periodically. That is the reason you get credits even if you haven't "finished" a WU. These aren't WU aren't like most projects where there is a beginning and an end. They are "dynamic" WU meaning they change the values they are computing.

This is the similar to the way that Climate Prediction works. They have very long WU that will actually finish at some point - sometimes months. However, while you are running a WU your machine sends information back to the project periodically and you get credits for the work completed to that point. These reports going back to the project are called "trickles".

Post Reply

Who is online

Users browsing this forum: No registered users and 1 guest