Index  | Recent Threads  | Unanswered Threads  | Who's Active  | Guidelines  | Search
 

Quick Go ยป
No member browsing this thread
Thread Status: Active
Total posts in this thread: 10
[ Jump to Last Post ]
Post new Thread
Author
Previous Thread This topic has been viewed 1379 times and has 9 replies Next Thread
nep nep
Cruncher
Joined: Apr 5, 2019
Post Count: 2
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
cluster processing

so my desktop runs WCG 24/7 however my laptops its fairyl random when i can run those so i dont particuly like to have them download tasks and have been looking to see if i can essentaly run a cluster mode where my laptop connects to my desktop and controbutes to some tasks that have not been started yet
however the select computer doesnt seem to connect and was hopeing someone here could point me in the correct direction for that or if i would need tot use a networked drive
[Jan 15, 2020 2:47:55 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: cluster processing

Chances of that succeeding in the present BOINC design are next to zero. You can run a minimal client, just fetch 1 task at the time, the smaller ones such as MIP1, MCM1 who have a 7 day deadline at that with <fetch_minimal_work>1</fetch_minimal_work> in the cc_config.xml file. Nothing will be fetched until the present task is completed and reported.
[Jan 15, 2020 3:00:15 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Mike.Gibson
Ace Cruncher
England
Joined: Aug 23, 2007
Post Count: 12362
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: cluster processing

Rather than using app_config, I would suggest setting Device Profiles on occasional use laptops to download only short computation long deadline projects such as mcm1 and mip1 and to limit the total number to the available threads.

The best would be mip1 as it has the longer deadline and shorter computation time.

Mike
[Jan 15, 2020 3:31:53 PM]   Link   Report threatening or abusive post: please login first  Go to top 
nep nep
Cruncher
Joined: Apr 5, 2019
Post Count: 2
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: cluster processing

ahh fair enough i kinda always assumed thats what the select pc thing was for
[Jan 15, 2020 5:37:50 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Mike.Gibson
Ace Cruncher
England
Joined: Aug 23, 2007
Post Count: 12362
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: cluster processing

I should have said that in Device Profiles you need to click on the Custom Profile radio button.

Mike
[Jan 15, 2020 6:29:28 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: cluster processing

Rather than using app_config, I would suggest setting Device Profiles on occasional use laptops to download only short computation long deadline projects such as mcm1 and mip1 and to limit the total number to the available threads.

The best would be mip1 as it has the longer deadline and shorter computation time.

Mike

Where is app_config written in my recommendation?
[Jan 15, 2020 7:41:23 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: cluster processing

ahh fair enough i kinda always assumed thats what the select pc thing was for

That selector allows the BOINC 'Manager' to connect to different client, local and remote, the remote a little involved needing messing with some config and hosts files. Many actually use BOINCTasks which is a multi-client manager. It even can connect to clients over the internet, but never tried that.
[Jan 15, 2020 7:47:38 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Mike.Gibson
Ace Cruncher
England
Joined: Aug 23, 2007
Post Count: 12362
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: cluster processing

lavaflow

You mentioned app_config but didn't actually recommend it. I was trying to suggest a simpler way to proceed.

Mike
[Jan 16, 2020 1:17:43 AM]   Link   Report threatening or abusive post: please login first  Go to top 
hchc
Veteran Cruncher
USA
Joined: Aug 15, 2006
Post Count: 798
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: cluster processing

I wish BOINC architecture allowed for clients to be able to download from a proxy serving middleware device we set up. That way we can micromanage work units after the fact; i.e. "you go here, you go there," but that functionality doesn't currently exist, and I wouldn't hold my breath it even lands on their backlog.
----------------------------------------
  • i5-7500 (Kaby Lake, 4C/4T) @ 3.4 GHz
  • i5-4590 (Haswell, 4C/4T) @ 3.3 GHz
  • i5-3570 (Broadwell, 4C/4T) @ 3.4 GHz

[Jan 16, 2020 12:01:42 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: cluster processing

I wish BOINC architecture allowed for clients to be able to download from a proxy serving middleware device we set up. That way we can micromanage work units after the fact; i.e. "you go here, you go there," but that functionality doesn't currently exist, and I wouldn't hold my breath it even lands on their backlog.

+1
I was hoping they would add an XML import/export function to allow a user to move a WU or WUs to another client. Since the client has it's shortcomings when downloading work that creates overloaded clients, it would be nice to be able to move some of that extra work to faster clients instead of watching it abort. There have also been times when one knows that a client machine is about to fail, so it would be nice to be able to move that work to other clients instead of just aborting the work. Once moved, the new machine would report the new location to the server on the next scheduler request and download any files that might be missing.
[Jan 16, 2020 12:14:07 PM]   Link   Report threatening or abusive post: please login first  Go to top 
[ Jump to Last Post ]
Post new Thread