I aborted the Atlas computation. Note that it logged the task as 'Finished' and Reported it. I am reluctant to take on new tasks. Other tasks run OK.

30/12/2014 15:17:30 | Cosmology@Home | Finished upload of wu_122814_125146_0_0_0_3
30/12/2014 15:17:30 | Cosmology@Home | Started upload of wu_122814_125146_0_0_0_4
30/12/2014 15:17:30 | Cosmology@Home | Started upload of wu_122814_125146_0_0_0_5
30/12/2014 15:17:32 | Cosmology@Home | Finished upload of wu_122814_125146_0_0_0_4
30/12/2014 15:17:32 | Cosmology@Home | Finished upload of wu_122814_125146_0_0_0_5
30/12/2014 15:32:38 | ATLAS@home | task qryNDmObmJlnDDn7oo6G73TpABFKDmABFKDm55GKDmABFKDmgaWUQo_0 aborted by user
30/12/2014 15:32:49 | ATLAS@home | Computation for task qryNDmObmJlnDDn7oo6G73TpABFKDmABFKDm55GKDmABFKDmgaWUQo_0 finished
30/12/2014 15:34:11 | ATLAS@home | Sending scheduler request: To report completed tasks.
30/12/2014 15:34:11 | ATLAS@home | Reporting 1 completed tasks
30/12/2014 15:34:11 | ATLAS@home | Not requesting tasks: "no new tasks" requested via Manager
30/12/2014 15:34:14 | ATLAS@home | Scheduler request completed
30/12/2014 16:18:25 | Cosmology@Home | Sending scheduler request: To report completed tasks.
30/12/2014 16:18:25 | Cosmology@Home | Reporting 1 completed tasks
30/12/2014 16:18:25 | Cosmology@Home | Not requesting tasks: don't need (not highest priority project)
30/12/2014 16:18:27 | Cosmology@Home | Scheduler request completed
30/12/2014 16:28:35 | Milkyway@Home | Starting task de_nbody_12_20_orphan_sim_2_1413455402_1530899_6
30/12/2014 17:01:19 | Milkyway@Home | Computation for task de_nbody_12_20_orphan_sim_2_1413455402_1530899_6 finished

Hello Gary & Chris,

I would like to add some comments about ATLAS.
I will try to be as polite as Jord.

I have done ATLAS WU last year (on Linux) - but not currently.
Atlas has reduced the size requir4ed to to run a WU.
I don't have the URL for the exact forum entry - but it is there.

I ended up using an app_config.xml to limit the number of concurrent Atlas WU.
That was one approach after I resized partitions and made a humonguous disk space available.

I also also had problems with the VM stalling.
I *think* I solved this by setting to task switch to 20 hours and
setting the checkpoint interval to about 20 minutes.
Before you do this, you may want to look at the checkpoint size and weigh
the decision to have less checkpointing.
I was paranoid about having 2 ATLAS task trying to checkpoint at the same time.
I have over 11GiB of memory and I tried running 4 Atlas WU at once - but ended up running only 2 at once.

Another consideration is the large IO done within the VM by ATLAS.
It took my medium-speed (USA standards) DSL about 20 minutes for the WU to satisfy itself. That pushed me to start a second WU about and hour after the first one.

Disclosure: I may be over 60, but I can act like a kid - wanting to see If I can run the project that uses a lot of memory and CPU. I run WCG now.

I hope this helps.
You may have already jumped over your hurdles.
(That is satisfying too.)
Please let me know how it turned out....

Thanks,
Jay

Tha Skynet Pogs also has plans to use a virtual machine.

Message http://pogs.theskynet.org/pogs/forum_thread.php?id=506&postid=3671#3671 says: I won't be able to explain this as well as kevin. How ever the short answer for using VBOX is the ability to use a greater number of science libraries. A lot of radio astronomy computing tools are written only for *nix systems and often require complicated setup processes to get them running. Thus they need to be deployed in a virtual box to allow them to run on the vast number of different machines that connect to BOINC.

We are well aware of the problems this causes to boinc clients that allready run on VMware and we are looking at what projects we would be able to run without using a VBox alongside the other projects.

Message http://pogs.theskynet.org/pogs/forum_thread.php?id=506&postid=3698#3698 says: Virtual Box

To get a good fit for the galaxies, we really would like to try many other codes (some of which only run on *nix). I’ll then use Bayesian Ensemble methods to improve the fit we get from the data by running multiple codes. These codes are written in a variety of languages including python, and they’re under constant development. I want to be able to just deploy the code and let it run, rather than having to spend precious time and resources converting updates constantly.

The Virtual Box version will allow me to do that as I can have all the libraries precompiled in the VB file and I download the latest executables (as sticky files so you only download the once) and run all of them against all the pixels you have downloaded. The VB will be an automatic process within BOINC, you won’t have to do anything extra, and it will mean more science bang for our bucks.

The stand alone executable will remain as we have a solid Android user base and they can’t use VB, so there will be an option for standalone processing for those that don’t want to use the VB option.