How to determine the optimised configuration for parallel

saj216 · June 5, 2020, 00:00

Hello guys,

I am using a cluster for a simulation. I wish to determine the optimal performance for my configuration. How do you in layman terms carry out a study to determine this. I have been advised to work out the completion time of 1000 timesteps with the adjustabletimestep feature off using different decompositions and subdomains. Is this the correct way to achieve this? My simulation is transient. Thanks.

Please describe in the correct study in the simplest terms thanks again

geth03 · June 10, 2020, 02:37

i can tell you what i do.

i run the same case with different numbers of processors with the scotch decomposition method and track the time needed to complete a certain amount of simulation time.

you can then plot the number of processors on the x-axis and the time needed on the y-axis. if you don't see any improvement after certain number of processors, i would say it is not neccessary to use more than that amount.

it is simple as that.

cheers

June 5, 2020, 00:00	How to determine the optimised configuration for parallel	#1
saj216 Member Sam Join Date: May 2019 Posts: 64 Rep Power: 6	Hello guys, I am using a cluster for a simulation. I wish to determine the optimal performance for my configuration. How do you in layman terms carry out a study to determine this. I have been advised to work out the completion time of 1000 timesteps with the adjustabletimestep feature off using different decompositions and subdomains. Is this the correct way to achieve this? My simulation is transient. Thanks. Please describe in the correct study in the simplest terms thanks again

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
[snappyHexMesh] snappyHexMesh error "Cannot determine normal vector from patches."	lethu	OpenFOAM Meshing & Mesh Conversion	1	June 3, 2020 07:49
[OpenFOAM.org] Cannot open configuration file mpicc-wrapper-data.txt mpicc: Not found	wenjuny	OpenFOAM Installation	0	November 25, 2019 09:01
need therm.dat file for ECFM3Z Combustion model	ankitraut27	CONVERGE	8	July 2, 2019 12:34
[OpenFOAM.com] Cannot open configuration file /opt/openmpi/share/openmpi/mpicc-wrapper-data.txt	lhsihan	OpenFOAM Installation	1	January 28, 2019 19:47
Multiple Configuration Simulation	Tristan	CFX	0	November 13, 2009 23:01

June 10, 2020, 02:37		#2
geth03 Senior Member Join Date: Dec 2019 Location: Cologne, Germany Posts: 355 Rep Power: 8	i can tell you what i do. i run the same case with different numbers of processors with the scotch decomposition method and track the time needed to complete a certain amount of simulation time. you can then plot the number of processors on the x-axis and the time needed on the y-axis. if you don't see any improvement after certain number of processors, i would say it is not neccessary to use more than that amount. it is simple as that. cheers