Once the Coalition Server is configured with available computational resources, the user can perform parallel computations by using Distributed Submission to submit processing tasks from the local workstation to the remote server, either through simple or batch workflow submission.
g-Platform allows most processing modules to be executed in distributed mode on CPU and GPU clusters. While not all workflows require distributed execution, it is essential for computationally intensive modules such as SRME 2D/3D, Radon multiple attenuation, Kirchhoff PreSTM and PreSDM, etc.
For execution any workflow/job in the distribution mode, we should go to the Settings tab of the particular module and choose Distribution mode with other parameters:
How to calculate Bulk Size parameter for migration:
GRT- is nothing but GPU RAM TOTAL;
For example, if RAM value is 11.2 GB, we will use only 10GB as 1GB of RAM is used by others processes;
GRA– is known for GPU RAM AVAILABLE;
For example, if RAM is 10 GB, the parameter value will be calculated as following:
10 * 1024 * 1024 * 1024 = 10737418240 (size in bytes);
Data parameters:
NS– indicates number of samples (for example, 2000 samples);
MinOFF– indicates the minimum offset of CIG (for example, 0m);
MaxOFF– indicates the maximum offset of CIG (for example, 10000m);
StepOFF– indicates the CIG offset step (for example, 200m);
TCIG– indicates the number of traces in one CIG, this parameter is calculated as following:
TCIG = (MaxOFF – MinOFF) / StepOFF, for example (10000-0)/200=50
CIG_B– indicates the size (in bytes) of a single CIG, this parameter is calculated as following:
TCIG * NS * 4, for example 50*2000*4=400000bytes
MAXIMUM BULK SIZE– is an optimal parameter for PSTM calculation, this parameter is calculated as following:
MAXIMUM BULK SIZE = GRA / CIG_B,for example 10737418240/400000=26843.5456 – this number indicates the maximum bulk size.
Please note, from common practice, it’s better to round this number down, use 26000.