Query regarding TraceWin remote computation: How to enable multi-core parallel computing in Error Study?

[Xu Fred]

Dear Didier,

I am using the Error Study in TraceWin and am connected to a high-performance computer for remote calculations.

I have successfully established a connection with the remote computer, the communication is working correctly, and I am able to run computations.

The issue I'm encountering is related to the parallel core settings. In the local TraceWin GUI, I have set the "Nbr of core" to 100, with the expectation of utilizing 100 cores on the remote machine for parallel processing. However, during execution, I've observed that only a single core (1 core) is being used for the computation on the remote machine. This makes the entire Error Study process very slow.

My goal is to leverage multiple cores on the remote computer (e.g., 100 cores) to run computations simultaneously, which would significantly reduce the total processing time.

Therefore, I would like to ask the following questions:

1.Does TraceWin support multi-core, distributed parallel computing on a remote machine when conducting an Error Study?

2.If it does, what are the correct configuration steps? It seems that simply setting the "Nbr of core" in the local GUI is not sufficient.

3.Do the script files on the remote machine, such as tw_job_run.sh, tw_job_status.sh, and tw_job_kill.sh, require specific parameter modifications to enable TraceWin to correctly submit multi-core jobs?

4. I'd like to clarify the job distribution logic for error studies (I think this question is based on that remote computer can use several cores for error study). With Nbr of steps = 5, Nbr of run = 10, and Nbr of cores = 20, how are the tasks assigned? Specifically, does TraceWin process all 10 runs for Step 1 before starting Step 2, or does it treat all 50 runs (5x10) as a single pool to be distributed among the 20 cores as they become available?

Thank you very much for your time and help!

Yours,
Xu

[Didier]

Dear Xu,

1 : Yes of course, it's the goal.
2 : Normally set "Nbr of core" is sufficient.
3 : These scripts are designed for clusters, not for multi-core machines.
4 : No, "Nbr of step" are processed sequentially, but normally "Nbr of step" should be much smaller than "Nbr of run" in all logic.

Send me an image of the "Remote computers" tool, I'll check in first step.

Regards,

Didier

[Xu Fred]

Dear Didier,

Thanks for your reply.

Perhaps my description was not clear enough. I am using a cluster to conduct an error study and would like to utilize the multi-core capabilities of the cluster for computation (e.g., 100 cores). I have configured tw_job_run.sh, tw_job_status.sh, and tw_job_kill.sh according to the TraceWin documentation (without modifying the parameters inside). Attached are some images.


Image 1-2 are setting before running the error study.

1.png (48.41 KiB) Viewed 3375 times

2.png (91.72 KiB) Viewed 3375 times

Image 3-1 and 3-2 are process of error study.

3-1.png (6.2 KiB) Viewed 3375 times

3-2.png (5.94 KiB) Viewed 3375 times

Image 4-7 are the information about remote computer when I run the error study.

4.png (103.59 KiB) Viewed 3375 times

5.png (102.19 KiB) Viewed 3375 times

6.png (101.31 KiB) Viewed 3375 times

7.png (101.46 KiB) Viewed 3375 times

Image 8 is another problem, and I do not know how to solve. By the way, what is "wanted memory(MByte)" mean in the remote computer setting in Image 2, I set a value of 20000.

8.png (18.68 KiB) Viewed 3375 times

Thank you for your time and help again.

Regards,
Xu

[Didier]

images are missing!

[Xu Fred]

images are missing!

I reloaded the images, please see the last reply.

[Didier]

ok thank for the images, I'll study that.
How are you install twserver, how do you run it and where exactly ?

[Didier]

Dear Xu,

There's an error in the text of the message, it's not "Remaining memory" but "Remaining disk space", that said, I don't understand this message which shouldn't arrive, you obviously have enough space on your cluster (see image# 1).

Could you answer my question about twserver, and try again using only two cores (it's good for testing) and when errors occur use the "Detail" button to read and view the errors.

Regards,

Didier