LHC@home

Hello everyone,

Wanted to mention something that unexpectedly happen when running this app.

I have a sixteen core system with only 24G of ram. (I have another 24G already ordered and will be coming in a few days.

My configuration was set to use 100% CPU's. This resulted in the server sending me sixteen virtual machines all of which was configured to use 2048M each. Thats 32G's which drew my workstation to a crawl. I managed to change the settings to use 75% CPU so it would only send 10 VM's. Once the other ram arrive I will set it back to 100% and see how the system performs.

Is there currently any way to use the ten VM's configured for two CPU's each with 2048G of ram? I opened Virtualbox and can see the ten VM's running. Should I stop each one and change the settings are should I just it run it's course until the new ram arrives?

Currently many of them end with computation errors.

Any advice would be greatly appreciated,

James

Sorry I meant run eight VM's with 2048GB so I would change my setting to 70% or 60% CPU's?

Currently many of them end with computation errors.

Any advice would be greatly appreciated ..

The LHCb sub-project is a beta application.
There are BOINC-tasks, but often there aren't jobs available for the VM.
If that's the case the task will end into an error after about 10 minutes.

To control the number of VM's/tasks you can limit the number either 2 ways:

- Set the max # jobs in your project preferences

- Use an app_config.xml in the lhcathome.cern.ch_lhcathome project directory.

Example of an app_config.xml. Change it to your needs.
If you only want to control the number of tasks, you can remove the app_version parts.

<app_config>
<project_max_concurrent>7</project_max_concurrent>
 <app>
  <name>ALICE</name>
  <max_concurrent>1</max_concurrent>
  <fraction_done_exact/>
 </app>
 <app>
  <name>ATLAS</name>
  <max_concurrent>1</max_concurrent>
  <fraction_done_exact/>
 </app>
 <app>
  <name>Benchmark</name>
  <max_concurrent>4</max_concurrent>
  <fraction_done_exact/>
 </app>
 <app>
  <name>CMS</name>
  <max_concurrent>4</max_concurrent>
  <fraction_done_exact/>
 </app>
 <app>
  <name>LHCb</name>
  <max_concurrent>1</max_concurrent>
  <fraction_done_exact/>
 </app>
 <app>
  <name>sixtrack</name>
  <max_concurrent>4</max_concurrent>
  <fraction_done_exact/>
 </app>
 <app>
  <name>Theory</name>
  <max_concurrent>4</max_concurrent>
  <fraction_done_exact/>
 </app>
 <app_version>
  <app_name>ALICE</app_name>
  <plan_class>vbox64</plan_class>
  <avg_ncpus>1.000000</avg_ncpus>
  <cmdline>--nthreads 1.000000</cmdline>
  <cmdline>--memory_size_mb 630</cmdline>
 </app_version>
 <app_version>
  <app_name>Benchmark</app_name>
  <plan_class>vbox64</plan_class>
  <avg_ncpus>1.000000</avg_ncpus>
  <cmdline>--nthreads 1.000000</cmdline>
  <cmdline>--memory_size_mb 1408</cmdline>
 </app_version>
 <app_version>
  <app_name>CMS</app_name>
  <plan_class>vbox64</plan_class>
  <avg_ncpus>1.000000</avg_ncpus>
  <cmdline>--nthreads 1.000000</cmdline>
  <cmdline>--memory_size_mb 1896</cmdline>
 </app_version>
 <app_version>
  <app_name>Theory</app_name>
  <plan_class>vbox64</plan_class>
  <avg_ncpus>1.000000</avg_ncpus>
  <cmdline>--nthreads 1.000000</cmdline>
  <cmdline>--memory_size_mb 630</cmdline>
 </app_version>
</app_config>

Wow, thats awesome Crystal.

So when the LHCb app is out of beta I can set the VM's to use four CPU's and 2048MB by changing the nthreads and memory_size_mb?

Will the LCHb app use more than one CPU with the VM? Seems it would comlete a lot faster you think?

James

Wow, thats awesome Crystal.

So when the LHCb app is out of beta I can set the VM's to use four CPU's and 2048MB by changing the nthreads and memory_size_mb?

Will the LCHb app use more than one CPU with the VM? Seems it would comlete a lot faster you think?

James

No it will not be faster, cause you will not run 1 job on 4 cores, but 4 jobs on 4 cores.
You have to increase the memory, but not necessary 4 times 2048MB. Maybe even 6GB could suffice.

One have to try and error for the best memory size to run more jobs simultaneously.

I don't run these particular tasks but running the Theory VB here (and over at vLHC) I have been able to run 8 tasks at the same time on my 8-core with only 8GB ram

Now with the CMS version those took a bit more ram so I only would run 5 or 6 and at time runnning the vLHC-dev X2 Theory tasks at the same time.

right now it is running vLHC-dev X2 (vbox64_mt_mcore) so that is 4 CPU cores and at the same time 3 LHC VB Theory tasks with the 8GB ram (I would like to plug another 4GB in that I admit)

My quads and a couple 3-core phenoms run ALL cores with the tasks here and they all have just 12GB ram each.

(I do have 3 new 8GB sticks of ram but just never get around to shutting one of mine down to plug em in so 3 of them will be 20GB.......which is more than enough for Cern crunching and Einstein GPU's)

Funny how I have these 7 desktops and if you add up all the CPU cores it is still less than one of those fancy new Xenons I see here now.

(actually not funny to me.......I will hate em until I get to have one here)

---

Volunteer Mad Scientist For Life

I agree about the new 22 core Xeon's Can't wait for them to be cheap on eBay.

What I meant on the four cores on one VM was to stop the VM and change the CPU count to four then start it back. Will it use all four cores?

Okay I see there is no way to stop the VM. Only suspend the work unit which puts the VM in saved mode. Everything is greyed out as far as the CPU count per VM.

I wanted to do this based on total RAM. 16 cores with 24GB of RAM.

It would have been nice to use four cores per VM's (work unit)that would have used 8GB and may have completed faster based on using all 16 cores.

I see about 55gb ram usage for my 16core Xeon running 32 VMs

On my 10 core machine with 32gb, I couldn't max the cpu as there wasn't enough ram, I think I could run about 16VMs