Maximum number of download
log in

Advanced search

Message boards : Number crunching : Maximum number of download

1 · 2 · Next
Author Message
Alessio Susi
Avatar
Send message
Joined: 7 Jun 15
Posts: 15
Credit: 661,981
RAC: 0
Italy
Message 1721 - Posted: 23 Mar 2020, 15:07:00 UTC

Hi. How can I set a maximum number of download for this project?
____________
ASUS X570 E-Gaming
AMD Ryzen 9 3950X, 16 core / 32 thread 4.4 GHz
AMD Radeon Sapphire RX 480 4GB Nitro+
Nvidia GTX 1080 Ti Gaming X Trio
4x16 GB Corsair Vengeance RGB 3466 MHz

Profile Buro87 [Lombardia]
Send message
Joined: 23 Nov 16
Posts: 100
Credit: 4,000,541
RAC: 0
Italy
Message 1722 - Posted: 23 Mar 2020, 18:39:44 UTC - in response to Message 1721.

Tn-grid has limit of 8wu per core, and 512wu per pc, even with a Epyc (64c/128t)
With your 3950x you can reach 32x8= 256 running wu

Aurum
Send message
Joined: 18 Jul 18
Posts: 97
Credit: 291,241,518
RAC: 308
United States
Message 1723 - Posted: 24 Mar 2020, 17:11:04 UTC

Alessio, Are you asking because you want more WUs or fewer???
If fewer, you can set the Resource to zero and then it should only send more when you run out.

Alessio Susi
Avatar
Send message
Joined: 7 Jun 15
Posts: 15
Credit: 661,981
RAC: 0
Italy
Message 1724 - Posted: 24 Mar 2020, 17:54:06 UTC

I have a Ryzen 9 3950X and I'm running 30 WUs simultaneously. I'd like to set TN-Grid to download only 20 tasks at a time. So I can run TN-Grid and other projects at the same time.
____________
ASUS X570 E-Gaming
AMD Ryzen 9 3950X, 16 core / 32 thread 4.4 GHz
AMD Radeon Sapphire RX 480 4GB Nitro+
Nvidia GTX 1080 Ti Gaming X Trio
4x16 GB Corsair Vengeance RGB 3466 MHz

Aurum
Send message
Joined: 18 Jul 18
Posts: 97
Credit: 291,241,518
RAC: 308
United States
Message 1725 - Posted: 25 Mar 2020, 2:16:05 UTC
Last modified: 25 Mar 2020, 2:19:46 UTC

Alessio, TN-Grid does not give us that kind of control like WCG does. But you can still make it work as you want using app_config.xml files in your project folders. Now I like to run both Rosetta and TN-Grid as my CPU projects.
In my gene.disi.unitn.it_test folder I have:

<app_config> <app> <name>gene_pcim</name> <!-- needs ? GB RAM per WU --> <!-- Xeon E5-2686 v4 18c36t 32 GB 45 MB L3 Cache --> <max_concurrent>25</max_concurrent> </app> </app_config>


In my boinc.bakerlab.org_rosetta folder I have:
<app_config> <app> <name>rosetta</name> <!-- needs 5 MB L3 cache per mip1 WU --> <!-- needs 1.5 GB RAM per WU --> <!-- Xeon E5-2686 v4 18c36t 32 GB 45 MB L3 Cache --> <max_concurrent>9</max_concurrent> </app> <app> <name>minirosetta</name> <!-- needs 5 MB L3 cache per mip1 WU --> <max_concurrent>1</max_concurrent> </app> <project_max_concurrent>9</project_max_concurrent> </app_config>

You'll get more WUs than you need but you'll run what you want.
My totals leave 2 CPU threads for GPUGrid WUs.
____________

Magiceye04
Send message
Joined: 14 Dec 16
Posts: 10
Credit: 3,461,526
RAC: 0
Germany
Message 2141 - Posted: 24 Dec 2020, 10:22:41 UTC - in response to Message 1722.

Tn-grid has limit of 8wu per core, and 512wu per pc, even with a Epyc (64c/128t)
With your 3950x you can reach 32x8= 256 running wu


Hello,
when do I get this 8 WU per Core?
I am running the project since some days now with more then 200 valid WUs but get only 1 WU per Core.
Its the same on 2 different PCs. The project was added completely new for both PCs.

Best Regards
Magiceye

Profile Buro87 [Lombardia]
Send message
Joined: 23 Nov 16
Posts: 100
Credit: 4,000,541
RAC: 0
Italy
Message 2142 - Posted: 24 Dec 2020, 10:53:22 UTC - in response to Message 2141.
Last modified: 24 Dec 2020, 11:14:38 UTC

Tn-grid has limit of 8wu per core, and 512wu per pc, even with a Epyc (64c/128t)
With your 3950x you can reach 32x8= 256 running wu


Hello,
when do I get this 8 WU per Core?
I am running the project since some days now with more then 200 valid WUs but get only 1 WU per Core.
Its the same on 2 different PCs. The project was added completely new for both PCs.

Best Regards
Magiceye



I think you need to set your work buffer from your boinc manager preferences
with yor 1700 and 3950x setting 1day + 1day, is more than enough to reach 128wu and 256wu respectively

You need also to considerer the max concurrent wu limit.
If you have set, for example; a limit of 10 concurrent wu, trough an app_config.xml file, you will recieve 10*8=80 wu even with a 10+10days work buffer

Magiceye04
Send message
Joined: 14 Dec 16
Posts: 10
Credit: 3,461,526
RAC: 0
Germany
Message 2144 - Posted: 24 Dec 2020, 16:39:47 UTC

The buffer is already set to 1 day. Einstein gets much more work.
The project was added new, no config files.

Magiceye04
Send message
Joined: 14 Dec 16
Posts: 10
Credit: 3,461,526
RAC: 0
Germany
Message 2145 - Posted: 24 Dec 2020, 20:43:06 UTC - in response to Message 2144.
Last modified: 24 Dec 2020, 20:49:14 UTC

I tried now 10+10 days and on the Ryzen 1700 I got now more WUs: 128 :)

But the 3950X still has only 28 WUs.
Do 24 Dez 2020 21:42:14 CET | TN-Grid Platform | Reporting 1 completed tasks
Do 24 Dez 2020 21:42:14 CET | TN-Grid Platform | Not requesting tasks: don't need (CPU: ; AMD/ATI GPU: )

the cc_config is quiet short:
<cc_config>
<log_flags>
<task>1</task>
<file_xfer>1</file_xfer>
<sched_ops>1</sched_ops>
</log_flags>
</cc_config>

Profile Buro87 [Lombardia]
Send message
Joined: 23 Nov 16
Posts: 100
Credit: 4,000,541
RAC: 0
Italy
Message 2146 - Posted: 24 Dec 2020, 22:24:47 UTC - in response to Message 2145.
Last modified: 24 Dec 2020, 22:31:09 UTC

Try to suspend others project, than ask work for TN-grid

If you want, the he following is the cc_config.xml i am using

<cc_config> <options> <allow_multiple_clients>1</allow_multiple_clients> <max_file_xfers>50</max_file_xfers> <max_file_xfers_per_project>30</max_file_xfers_per_project> <report_results_immediately>1</report_results_immediately> <ncpus>12</ncpus> </options> </cc_config>


set the ncpus value to your correct thread number: 32 (3950x), 16 (1700)

Others values are for multiple wu trasfer


By the way i write down also the app_config.xml file to put in the project folder (C:\ProgramData\BOINC\projects\gene.disi.unitn.it_test)
whit this you can specify the maximum number of concurrent wus. It's useful if you want run multiple projects

<app_config> <project_max_concurrent>12</project_max_concurrent> </app_config>

Magiceye04
Send message
Joined: 14 Dec 16
Posts: 10
Credit: 3,461,526
RAC: 0
Germany
Message 2147 - Posted: 25 Dec 2020, 0:01:55 UTC
Last modified: 25 Dec 2020, 0:02:53 UTC

I have no other projects running for the CPU at the moment that I could suspend.
Einstein is using the GPU. If I suspend Einstein, nothing changes.

I dont want to set any fix limit for CPU cores. Such limits might be the reason for the problems like I have - so I avoid this wherever possible.
The 1700 got some more work now with setting of 10+10 days. But switching back to 1+1 it doesn't request new work again.

The 3950X still has only 1 WU per core even with 10+10 days setting.

If there are no other ideas I will ignore the problem, because other projects run fine and TNgrid also, as long as there are always enough WUs available.

Profile Keith Myers
Send message
Joined: 26 Jun 20
Posts: 63
Credit: 15,280,917
RAC: 30
United States
Message 2148 - Posted: 25 Dec 2020, 0:02:07 UTC

To get a fully populated cc_config.xml instead of the minimal default, all you have to do is change ONE Event Log Options parameter in the Manager menu and Save the setting.

That will populate all the possible settings of the cc_config.xml file.

You can then change the one setting you adjusted back to whatever it was before and it will keep the full set of config file parameters.

Magiceye04
Send message
Joined: 14 Dec 16
Posts: 10
Credit: 3,461,526
RAC: 0
Germany
Message 2151 - Posted: 28 Dec 2020, 11:33:10 UTC

I added TNgrid new on the fresh installation of Ubuntu and Boinc on another PC.
This PC also has the same problem. Only one WU per Core. Even after more than 20 valid WUs crunched.

For me this means, it has nothing to do with the PCs itself, no bad settings somewhere, its just project related.

Profile Keith Myers
Send message
Joined: 26 Jun 20
Posts: 63
Credit: 15,280,917
RAC: 30
United States
Message 2154 - Posted: 29 Dec 2020, 6:22:55 UTC

You've got a client limitation being applied through the Manager likely.

If I wanted I could have all of my 48 cpu cores crunching TN-Grid and over 500 tasks in the cache.

Look for storage limits, task limits being set on a different venue than you think the hosts are set for.

Magiceye04
Send message
Joined: 14 Dec 16
Posts: 10
Credit: 3,461,526
RAC: 0
Germany
Message 2158 - Posted: 30 Dec 2020, 14:07:48 UTC
Last modified: 30 Dec 2020, 14:13:05 UTC

The only difference I see in the settings is that one PC is in the SCHOOL profile and the other two have no profile.
I have moved one of the PCs now also to SCHOOL, but no change. SCHOOL and standard settings are identical anyway.

Maybe that its also a limit from client side in conjunction with the project - but I have installed Boinc, added the projects and thats it. I have not limited anywhere the number of cores to be used. And even with 10+10days there is not a single WU in the buffer. 4 WUs for a 4core CPU.

Magiceye04
Send message
Joined: 14 Dec 16
Posts: 10
Credit: 3,461,526
RAC: 0
Germany
Message 2159 - Posted: 30 Dec 2020, 14:40:40 UTC - in response to Message 2158.

Wow, a wonder happened. The Athlon got now 7 additional WUs (for the next 20 days...).

So this means, the project needs a profile to be set for a PC and without a profile only 1 WU per Core is send out?

rlpm
Send message
Joined: 24 Mar 20
Posts: 4
Credit: 3,221,439
RAC: 0
United States
Message 2160 - Posted: 30 Dec 2020, 15:25:18 UTC - in response to Message 2159.

Wow, a wonder happened. The Athlon got now 7 additional WUs (for the next 20 days...).

So this means, the project needs a profile to be set for a PC and without a profile only 1 WU per Core is send out?

I think the work generator isn't keeping up with all the hosts requesting tasks? See here.

Magiceye04
Send message
Joined: 14 Dec 16
Posts: 10
Credit: 3,461,526
RAC: 0
Germany
Message 2162 - Posted: 30 Dec 2020, 16:03:10 UTC - in response to Message 2160.

Wow, a wonder happened. The Athlon got now 7 additional WUs (for the next 20 days...).

So this means, the project needs a profile to be set for a PC and without a profile only 1 WU per Core is send out?

I think the work generator isn't keeping up with all the hosts requesting tasks? See here.


No, I have 3 PCs running and only 2 of them had the problem over the last 2 weeks, even at the time where the work generator was fine.
And the one PC where I changed the location/profile has now enough work in the buffer.

Bryn Mawr
Send message
Joined: 23 Jun 20
Posts: 43
Credit: 14,256,442
RAC: 8
United Kingdom
Message 2163 - Posted: 30 Dec 2020, 17:36:05 UTC - in response to Message 2162.

Wow, a wonder happened. The Athlon got now 7 additional WUs (for the next 20 days...).

So this means, the project needs a profile to be set for a PC and without a profile only 1 WU per Core is send out?

I think the work generator isn't keeping up with all the hosts requesting tasks? See here.


No, I have 3 PCs running and only 2 of them had the problem over the last 2 weeks, even at the time where the work generator was fine.
And the one PC where I changed the location/profile has now enough work in the buffer.


Then you need to examine your default profile very closely because it is showing non-standard behaviour

p3d-cluster
Send message
Joined: 2 Apr 20
Posts: 3
Credit: 34,139,064
RAC: 0
Germany
Message 2164 - Posted: 30 Dec 2020, 17:55:56 UTC - in response to Message 2163.
Last modified: 30 Dec 2020, 18:05:42 UTC

Could you be so kind and have a look at my system 65300, please?
Why do i get only 52 WUs with my 256 Threads?

I'vent any app_config.xml, cc_config.xml or app_info.xml.

I use the school-profil with his default-setting.

Same on system 65301, 65302 and 65303.

Only TN-Grid is running.

1 · 2 · Next
Post to thread

Message boards : Number crunching : Maximum number of download


Main page · Your account · Message boards


Copyright © 2024 CNR-TN & UniTN