- #1
Vigardo
- 88
- 7
Hello experts,
Is it possible to run several instances of Ansys Mechanical in Windows or Linux in parallel? How?
I need to dramatically speed up the solution of a large list of Ansys Mechanical jobs, typically more than 50000. Such analyses are performed by some script that runs each job in batch mode.
First, I tried to vary the number of processors from 1 to 4 with the "-np" option but the acceleration was very small, i.e. 10-20% only. This probably is caused by the small size of my system, typical execution time per job is between 1 and 10 seconds.
Second, as a workaround, I tried to split up the job list in two and run each part in different terminals, but unfortunately Ansys crashes.
This was the command I used in this test under Windows, is it ok?
#> "<MY_ANSYS_PATH>\ansys145.exe" -p ansys -dir <MY_WORK_DIRECTORY> -j [JOB_NAME] -s noread -l en-us -b nolist -np 2 -i <INPUT_FILE> -o <OUTPUT_FILE>
Third, I also tried to run two different jobs at the same time in different terminals that used a different working directory. Unforutnately one of them crashes first (windows alerts that Ansys stopped working) while the other finishes normally later. Note that to avoid Ansys-lock-related problems I set the Ansys´ environment variable ANSYS_LOCK to OFF (Ansys manual says that by default ANSYS_LOCK is ON "to create file locks that prevent users from opening a new job with the same name and in the same directory as the current job").
These were the new windows commands tested:
On_Terminal_1#> "<MY_ANSYS_PATH>\ANSYS145.exe" -p ansys -dir "<WORK_DIR_A>" -j job_A -s noread -l en-us -b nolist -np 1 -m 512 -db 256 -i my_job_A.txt -o my_job_A.out
On_Terminal_2#> "<MY_ANSYS_PATH>\ANSYS145.exe" -p ansys -dir "<WORK_DIR_B>" -j job_B -s noread -l en-us -b nolist -np 1 -m 512 -db 256 -i my_job_B.txt -o my_job_B.out
Any suggestion? I really don´t know what to do next... Thanks!
Is it possible to run several instances of Ansys Mechanical in Windows or Linux in parallel? How?
I need to dramatically speed up the solution of a large list of Ansys Mechanical jobs, typically more than 50000. Such analyses are performed by some script that runs each job in batch mode.
First, I tried to vary the number of processors from 1 to 4 with the "-np" option but the acceleration was very small, i.e. 10-20% only. This probably is caused by the small size of my system, typical execution time per job is between 1 and 10 seconds.
Second, as a workaround, I tried to split up the job list in two and run each part in different terminals, but unfortunately Ansys crashes.
This was the command I used in this test under Windows, is it ok?
#> "<MY_ANSYS_PATH>\ansys145.exe" -p ansys -dir <MY_WORK_DIRECTORY> -j [JOB_NAME] -s noread -l en-us -b nolist -np 2 -i <INPUT_FILE> -o <OUTPUT_FILE>
Third, I also tried to run two different jobs at the same time in different terminals that used a different working directory. Unforutnately one of them crashes first (windows alerts that Ansys stopped working) while the other finishes normally later. Note that to avoid Ansys-lock-related problems I set the Ansys´ environment variable ANSYS_LOCK to OFF (Ansys manual says that by default ANSYS_LOCK is ON "to create file locks that prevent users from opening a new job with the same name and in the same directory as the current job").
These were the new windows commands tested:
On_Terminal_1#> "<MY_ANSYS_PATH>\ANSYS145.exe" -p ansys -dir "<WORK_DIR_A>" -j job_A -s noread -l en-us -b nolist -np 1 -m 512 -db 256 -i my_job_A.txt -o my_job_A.out
On_Terminal_2#> "<MY_ANSYS_PATH>\ANSYS145.exe" -p ansys -dir "<WORK_DIR_B>" -j job_B -s noread -l en-us -b nolist -np 1 -m 512 -db 256 -i my_job_B.txt -o my_job_B.out
Any suggestion? I really don´t know what to do next... Thanks!