Hello everyone,

The following are the new runs for N-Body:

de_nbody_02_27_2023_v182_pal5__data__1
de_nbody_02_27_2023_v182_pal5__data__2
de_nbody_02_27_2023_v182_pal5__data__3

These are our first runs of Palomar 5 with data from SDSS DR17.
Please let me know if there are any issues with these new runs.

Thanks,
Kevin

We are seeing that the runs start at about an hour of runtime and after 4 hours the runtimes increase to more than 3 days. Is anybody seeing this?

Yes, the Elapsed Time and Remaining Time just keep on increasing. Restarting Boinc helps, but the times reported are still not meaningful. Many jobs take days rather than hours.

Yes, the Elapsed Time and Remaining Time just keep on increasing. Restarting Boinc helps, but the times reported are still not meaningful. Many jobs take days rather than hours.

eg
Milkyway@Home 8.231% Running (8 CPUs) 09:26:25 4d 09:15:03 7/04/23 3:11:46 PM Milkyway@Home N-Boby Simulation 1.82 (mt)
and Time Remaining is climbing faster than Time Elapsed is going down. Progress% is not changing.[/img]

Yes, the Elapsed Time and Remaining Time just keep on increasing. Restarting Boinc helps, but the times reported are still not meaningful. Many jobs take days rather than hours.

If they are using CPU time, just let them run. Estimates for nBody were never accurate, in particular after starting new runs.

---

Here is a workunit that is causing issues...
https://milkyway.cs.rpi.edu/milkyway/workunit.php?wuid=829552226

name	de_nbody_02_27_2023_v182_pal5__data__3_1674667492_1085804
application	Milkyway@home N-Body Simulation
created	13 Mar 2023, 20:52:13 UTC

All tasks are for Windows. It ran normally for one person, but has caused "error while computing" or "timed out" for everyone else. Mine is the abort because the task said it would take 12 more days to finish (past deadline). Notice how, for my aborted task, the runtime is over 10 times larger than CPU time! I assume that the calculations had halted long before I aborted it. I was running 4 cores. From the looks of the successful task, they were running 10 cores (researching their i9-10900T CPU confirms 10 cores).

I have about 50 N-body Simulation’s tasks with estimated run time of 20-50 min. per task. However, in fact, the tasks took 3-8 hours to complete runs. (A month ago I noticed some of tasks ran more than a day, while the estimated run time were only of 4-5 hours!). With this situation, many of my tasks at the lated queues will be aborted because they hit deadline.
I suggest the run time of N-body tasks should be estimated to 5-10-fold increase. In that case less tasks will be downloaded, so all of the downloaded tasks could complete their runs in due time.

I have about 50 N-body Simulation’s tasks with estimated run time of 20-50 min. per task. However, in fact, the tasks took 3-8 hours to complete runs. (A month ago I noticed some of tasks ran more than a day, while the estimated run time were only of 4-5 hours!). With this situation, many of my tasks at the lated queues will be aborted because they hit deadline.
I suggest the run time of N-body tasks should be estimated to 5-10-fold increase. In that case less tasks will be downloaded, so all of the downloaded tasks could complete their runs in due time.

I'm not sure it works that way, when they make the tasks they estimate how long it might take add a fudge factor and then send them out for us to crunch, adjusting the time now is up to the Boinc software and it can take more than 10 valid tasks in a row to even begin to get an idea of how long they will take and then ask the scheduler for new work based on that estimate. Tom the old Admin said that when they release new runs in the beginning there is ALOT of data to analyze and therefore the tasks can take longer while at the end of the runs there is less data so they are much faster. He said to think of it as a snapshot of a portion of the MilkyWay Galaxy, in the beginning all you see is a ton of Stars but as you begin to zoom in the Stars start to spread out and the you see more and more clear space between them.

... but as you begin to zoom in the Stars start to spread out and the you see more and more clear space between them.

Well, supposedly ...
All I can see at the moment is heavy slow fog ...

Nice Sunday to all !

cheers
S-F-V

Hello, I just install the latest mac OS Ventura 13.3.1 ... and Milky Way just do nothing !?! Help. I throw away Boing and re -download it ... with no effect ??? Marc, Switzerland.

Hello, I just install the latest mac OS Ventura 13.3.1 ... and Milky Way just do nothing !?! Help. I throw away Boing and re -download it ... with no effect ??? Marc, Switzerland.

Double posting is useless.
You are here in the N-Body section.
But as I see it, you are having troubles with Separation.

Maybe you are better off (??) posting under
"Platform-specific problems" --> "Macintosh"
wich might raise more interest ?

Have you read the STDERR output ?
Do you have newest drivers and necessary new functions ?

Just a thought of mine --------------
Have a nice day
S-F-V

Hello, I just install the latest mac OS Ventura 13.3.1 ... and Milky Way just do nothing !?! Help. I throw away Boing and re -download it ... with no effect ??? Marc, Switzerland.

Well your device with the M2 cpu is working just fine and running 8 tasks right now, but your other device is not getting any tasks, go thru your settings to ensure they are similar to your other pc and see if you can find something different between them, that may help you come back to the Forums with more details.

Most of these runs freeze at some point. If you suspend them for a little time, they will resume computation, but most likely you will need to suspend them again after some time.
Another thing I noticed, the progress bar can rise quickly to 40-50% at the beginning, but then it resets to 0% and after that starts rising very slowly.

Most of these runs freeze at some point. If you suspend them for a little time, they will resume computation, but most likely you will need to suspend them again after some time.
Another thing I noticed, the progress bar can rise quickly to 40-50% at the beginning, but then it resets to 0% and after that starts rising very slowly.

Are you running the nbody tasks with the default setting of using all available cpu cores? If so that could be your problem, try using an app_config file to run multiple tasks at the same time but only using 2 or 3 cpu cores per task.

Most of these runs freeze at some point.

Do you let them use 100% of CPU time? I had that issue with nBody when I tried running them on my laptop and get the noise to some acceptable level by limiting the CPU time used.

---

have also about 20% failure or some bad ,,,

have also about 20% failure or some bad ,,,

that seems high:
In progress (45) · Validation pending (4) · Validation inconclusive (62) · Valid (85) · Invalid (2)

That is my n-body stats as of right now.

I'm using 90% of the CPU's and 75% of CPU time. My oldish PC has 8 cores and starts making annoying noise if I push it too hard.
Status of the tasks says they are using 6 CPU's at the moment. Could that be the problem?
And where can I find more information about this app_config file, that sounds like something I want to check out later when I have more time.

I'm using 90% of the CPU's and 75% of CPU time. My oldish PC has 8 cores and starts making annoying noise if I push it too hard.

Use less cores, for example 50% and 100% of CPU time and the issue with stuck Wus will likely disappear.

If you need an app_config.xml, this one will run one nBody on 4 cores (you should still set CPU cores to 50% in BOINC Manager):

<app_config>
 <app>
  <name>milkyway_nbody</name>
  <max_concurrent>1</max_concurrent>
 </app>
 <app_version>
  <app_name>milkyway_nbody</app_name>
  <plan_class>mt</plan_class>
  <avg_ncpus>4</avg_ncpus>
  <cmdline>--nthreads 4</cmdline>
 </app_version>
</app_config>

---

Marko:

.....
And where can I find more information about this app_config file, that sounds like something I want to check out later when I have more time.

I'll try to show you how to find informations about the app_config parameter:

---- 1. In BOINC Manager advanced screen go to the "Help" tab
------- 2. In the dropdown menu select "BOINC help"
---------- 3. In the new browser window on the left side under "resources" select "User Manual"
------------- 4. In the next window under "Running BOINC: advanced" select "Client configuration"
---------------- 5. In the next window further down select "Project-level configuration"
------------------- 6.There you should find a desciption for the parm named app_config
---------------------- 7. If not, then I messed up these instructions !

Have a nice day
S-F-V

(I'm sure someone will point out an easier way to get there)