Einstein@home on dedicated HPC cluster

Azat
Azat
Joined: 12 Feb 18
Posts: 3
Credit: 10332630
RAC: 0
Topic 213367

Hi everyone,

I am newbie here and want to get some information about the project itself Smile

I will start by telling why I am actually posting here and interested in the project. For my research I need scientific applications that are usually running for a long period of time on GPU (>2 hours) in the HPC cluster. And I need to run these applications in order to get the data I want, I am not interested in the output of the application itself, but only in the GPU measurements (level of GPU/memory utilization, power consumtion, etc.). But I also want to run some meaningful application in order to not waste resources of the cluster. And this project is perfect for my purposes. 

So my question is, do you use BOINC only or do you have a dedicated HPC cluster where you run the computations using another software. If you have a cluster version, how much does it differ from BOINC one? 

I would appreciate any help, links or source code which even partially address my question. Have a nice day, thank you

Best regards,
Azat

Shawn Kwang
Shawn Kwang
Joined: 3 Nov 15
Posts: 289
Credit: 2741270
RAC: 1030

Azat, There are people who

Azat,

There are people who run E@H on HPC clusters, for example at AEI on the ATLAS cluster. They use boinc via the command line. (As BOINC is the only way to run E@H.) A dedicated master computer can control a number of boinc clients too. Some of our power-users do this.

To that end I'm moving your question to the Cruncher's Corner section of the forums.

 

Einstein@Home Project

Azat
Azat
Joined: 12 Feb 18
Posts: 3
Credit: 10332630
RAC: 0

Hi Shawn, Thank you very

Hi Shawn,

Thank you very much for the fast reply. It sounds very promising then. Might be you could help me further. Especially with that part:

As BOINC is the only way to run E@H

Is that stated somewhere, maybe in some papers? Wherever I can reference or cite? Would be really helpful. And thank you once again for your time. Have a good day! 

solling2
solling2
Joined: 20 Nov 14
Posts: 219
Credit: 1563279962
RAC: 47350

Azat_4 schrieb:... I would

Azat_4 wrote:

... I would appreciate any help, links or source code which even partially address my question. ...

I'd recommend the link on top of this page: science - publications. A lot to read there! :-)

It seems to me that the Einstein project is more adjusted to high throughput than to high performance. Have you got any early numbers how fast your special cluster can do a task?

 

Azat
Azat
Joined: 12 Feb 18
Posts: 3
Credit: 10332630
RAC: 0

solling2 wrote: I'd

solling2 wrote:

I'd recommend the link on top of this page: science - publications. A lot to read there! :-)

Thank you, I have found some information about ATLAS there as was mentioned in the previous answer. It might be helpful. Actually it might be sufficient for me. 

solling2 wrote:

Have you got any early numbers how fast your special cluster can do a task? 

It is not a special one :) Here it is: GWDG HPC

And I cannot tell how well it perforemed, because don't know what to compare with. But here is some of performance on tasks https://einsteinathome.org/host/12625659/tasks/4/0. Maybe you know better how to evaluate it and share with me ;)

solling2
solling2
Joined: 20 Nov 14
Posts: 219
Credit: 1563279962
RAC: 47350

Azat_4 schrieb: solling2

Azat_4 wrote:

solling2 wrote:

Have you got any early numbers how fast your special cluster can do a task? 

It is not a special one :) Here it is: GWDG HPC

Ah, man-made components ! I was hoping you run a special magic secret quantum computer. ;-) Meanwhile though I learned that quantum computers can't possibly work, even in principle: https://www.quantamagazine.org/gil-kalais-argument-against-quantum-computers-20180207/

For your studies on power consumption, you may find this discussion interesting, for comparison to your components: https://einsteinathome.org/de/content/radeon-vega

For your studies of GPU/memory utilization, just keep in mind that you have a better GPU load when allowing to run more than one task at a time. Anyway, good luck to your research!

 

Mumak
Joined: 26 Feb 13
Posts: 325
Credit: 3292961077
RAC: 1511116

700 s per GPU task is a bit

700 s per GPU task is a bit slower than a Radeon RX 480, here such one for comparison: https://einsteinathome.org/host/7192129/tasks/0/0

 

-----

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.