Development for LHC@home

Have put the limit to 5 tasks in progress.

Yep, thats about the limit for a 16GB host.

Aww, but what about my 128 GB host (with 20 cores...)?

It seems to have a terrible performance:
State: All (66) · In progress (0) · Validation pending (0) · Validation inconclusive (0) · Valid (16) · Invalid (0) · Error (50)

Have put the limit to 5 tasks in progress.

Yep, thats about the limit for a 16GB host.

Aww, but what about my 128 GB host (with 20 cores...)?

It seems to have a terrible performance:
State: All (66) · In progress (0) · Validation pending (0) · Validation inconclusive (0) · Valid (16) · Invalid (0) · Error (50)

That's probably due to the Fibre internet connection he has.
Shame the Fibre in question is a piece of damp string ;-)

There were no jobs available when I fired it up just now; looks like I've downloaded ten since.
[Edit] Ah no, five. The other five are on my Linux desktop. [/Edit]

---

That's probably due to the Fibre internet connection he has.
Shame the Fibre in question is a piece of damp string ;-)

I've seen it actually. On the roof of the Howell Centre you can just see a Heinz Soup can, with a long piece of string that runs to one of those tiny top-floor windows of Tower D.

Have put the limit to 5 tasks in progress.

Yep, thats about the limit for a 16GB host.

Aww, but what about my 128 GB host (with 20 cores...)?

Easy! In the early days T4T doesn't want to overload volunteer machines and restricted 1 task per host and the 32bit-Linux VM only used 256MB.
Now it's limited to 2 per host. I created multiple BOINC-clients to start for my needs to separate projects, subprojects, limiting resource share problems etc.

My AMD A10-6700 has 4 cores. I have now 4 vLHCathome-dev tasks running, one is a CMS, two are Theory 0.01 and one LHCb 0.05. Only GPU tasks, both SETI@home and Einstein@home run beyond the 4 cores limit.
Tullio

---

I tried creating an app_config.xml file in the projects/lhcathomedev.cern.ch_vLHCathome-dev directory with the following content.

<app_config>
  <project_max_concurrent>1</project_max_concurrent>
</app_config>

However my client still get 5 tasks. What I am doing wrong?

To answer my own question, this attribute is new in 7.4.9 and I have 7.2.42

https://boinc.berkeley.edu/wiki/Client_configuration#Application_configuration

From what I can see, I can add this file for the project so that novice users will only get one task but advanced users can find it and tune it. Suggestions welcome for a good default version.

When a user requests work, he requests a certain amount of seconds.
I believe, the server has a wrong setting, not knowing how long the tasks are.
The server needs to be told, that a task has a lenght of x seconds.
That way, boinc will only download the number of tasks with a total lenght,specified in the work buffer settings.
Currently, that is not the case.

Thanks any idea what and where that setting is on the sever?

My guess: estimated computation size (GFlops).For a tasks.

The current time remaining of an unstarted task is 7h38min (on my computer)
This value will be corrected gradually, the more tasks i am doing.
However, it helps, if it is close to the real value.(more like 14h, in my case)


There must be other poeple, who know more about boinc server settings.

Also, if the "project" max behaves as the "app" max does, you may still download tasks up to the number of available cores, but only the number specified will run, the others will wait.

Thanks any idea what and where that setting is on the sever?

You might get some ideas from here

Thanks any idea what and where that setting is on the sever?

You might get some ideas from here

"Last modified 5 years ago" -- I think CreditScrew is more recent than that, isn't it?

---

The task properties show estimated 1,000,000GFLPOPs.
I would set that at least to 50,000,000 GFLOPs. EDIT: would not help
There is still a chance, that some other setting like (max tasks) will disable that function.

Thanks any idea what and where that setting is on the sever?

You might get some ideas from here

"Last modified 5 years ago" -- I think CreditScrew is more recent than that, isn't it?

Yes but this is about job size estimation rather than credit.
The credit stuff is here

I found this:

https://boinc.berkeley.edu/boinc.pdf

It is a bit old, but it still applies (i think)

Chapter 11 and 12 is quite interesting.

Can you make it not over write the appconfig frequently?

Please see message from Crystal Pellet:

Yes, cause from BOINC client configuration (cc_config.xml in BOINC Data-directory):

<dont_check_file_sizes>0|1</dont_check_file_sizes>
Normally, the size of application and input files are compared with the project-supplied values after the files are downloaded and just before starting an application.
If this flag is set, this check is skipped. Use it if you need to modify files locally for some reason.

I restarted BOINC and had no problems with my modified and extended app_config.xml. Project size 411 bytes, my size 594 bytes.
But I already had the flag <dont_check_file_sizes> set to true (1) in my cc_config.xml.
Yes, cause from BOINC client configuration (cc_config.xml in BOINC Data-directory):

<dont_check_file_sizes>0|1</dont_check_file_sizes>
Normally, the size of application and input files are compared with the project-supplied values after the files are downloaded and just before starting an application.
If this flag is set, this check is skipped. Use it if you need to modify files locally for some reason.

I restarted BOINC and had no problems with my modified and extended app_config.xml. Project size 411 bytes, my size 594 bytes.
But I already had the flag <dont_check_file_sizes> set to true (1) in my cc_config.xml.