-
-
How do I do parallel processing on a cluster?
The Wavefront Batch Tracer can be executed in parallel on a MPP (Massively Parallel Processor) architecture, a cluster architecture or a network of workstations, e.g. Linux PCs.
The Wavefront Master Batch Tracer will divide the simulation into a number of separate, independent jobs, each consisting of one or more shots. One job runs on one node.
Each node runs an instance of the NORSAR-3D Batch Tracer module and the node must therefore have access to the software disc, project data disc and must be able to communicate with the license server.
The license server can be any machine in the network as long as it is reachable by all nodes and machines running NORSAR-3D. -
How can I use ssh instead of rsh?
On some systems (for example clusters) it is not possible to reach a node with a 'remote shell login' (rsh) command. In these cases the system administrator can configure the batch tracer by editing the user defined script rsh-script. The script is used to execute a command on a remote machine.
Edit the script <installed_n3d_path>/bin/scripts_user/TEMPLATE_rsh to use ssh instead of rsh. The filename must be on the form xxxx_rsh.sh (for example mycluster_rsh.sh) and it must be located at <installed_n3d_path)/bin/scripts_user
See the <installed_n3d_path>/bin/scripts_user/TEMPLATE_rsh script for a detailed description.To use the script in the batch tracer, do the following:
- Open the 'Machines Setup' window in the Batch Master and select ‘Machines…’
- Select the remote machine
- Select 'Use Advanced Configuration'
- Select 'Rsh Script' and push the arrow button
- Select the xxxx_rsh.sh script (for example mycluster_rsh.sh)
- Select OK and then OK in the ‘Set Machines ‘ window. You will now see that the script is listed in the machine list.
Please note that the 'Single Access Check' and the 'Initial Check Access' are not supported for machines that use Advanced Configuration.
In addition you must setup a password free ssh login for your user:- make an empty passphrase with the command: ssh-keygen –t dsa
- append key to authorized keys: cat .ssh/id_dsa.pub >> .ssh/authorized_keys
- update known_hosts for each node: ssh user@hostXX
-
How can I configure the batch tracer to run on a cluster?
On some systems (for example clusters) it is not possible to reach a node with a 'remote shell login' (rsh) command or you want to use a queing system on the cluster. In these cases the system administrator can configure the batch tracer by editing two user defined scripts: the rsh-script and the job-script. The rsh-script is used to start the batch remotely on another machine. (See "How can I use ssh instead of rsh?" above.) The job-script is used to start the slave job on the remote machine.
Job Script:
This script is used to start the job on the remote machine. See the <installed_n3d_path>/bin/scripts_user/TEMPLATE_job script for a detailed description. The filename must be on the form xxxx_job.sh and it must be located in <installed_n3d_path)/bin/scripts_user -
How can I run the batch tracer with a job scheduler?
If you want to use a job scheduler system on the cluster, the system administrator can configure the batch tracer by editing a user defined job-script. The job-script is used to start the slave job on the remote machine.
- Configuring the job-script
- Make a copy of <installed_n3d_path>/bin/scripts_user/TEMPLATE_job.sh
- Your copy must be located on the same directory as the template.
- Edit the last line in the script to this (MOAB example): qsub [ qsub_option(s) ] ${CMD} ${MODULE}
- Configure the batch tracer to use the script
- Start the WaveFront Batch Tracer
- Open the 'Set Machines' window in the Batch tracer
- Select the remote machine
- Select 'Use Advanced Configuration'
- Select 'Job Script' and push the arrow button
- Select your job script
- Close the window and push 'Change Machine' and then OK.
The 'Single Access Check' and the 'Initial Check Access' are not supported for machines which use Advanced Configuration.
- Configuring the job-script
-
Map Utility
-
How do I load wells into NORSAR-3D and show them in Map Utility?
The Map Utility can import wells on the NORSAR SwtrD file format (see below.) The file must have suffix '.SwtrD'
Well example file:# well_trajectory 1 * X,Y : Absolute UTM x,y at surface location in km * dx,dy : Relative UTM x,y to surface location in km * TVD : True Vertical Depth in km * DAH : Depth Along Hole(Measured Depth) in km ** X Y TVD 1000.000 2000.000 0.0 * dx dy TVD DAH 0.0 0.0 0.5 0.5 0.0 0.0 1.0 1.0 ...
-
NORSAR-3D and PSDM
-
Some guidelines for use of NORSAR-3D in PSDM applications
NORSAR-3D is a powerful tool for calculation of Green's functions, that is, tables of traveltimes and other parameters, for Kirchhoff type Pre-Stack Depth Migration.
In the following we discuss some aspects of the use of NORSAR-3D in the PSDM context. A general knowledge of NORSAR-3D is assumed.
The subsurface model
In NORSAR-3D the subsurface model can be a single velocity field, provided as a 3D (Seg-Y) cube, or a layered model with velocity discontinuities across the interfaces, or any combination of these. Notice that inside each block/volume in the model, the velocities must vary slowly. You can filter (smooth) velocity fields in NORSAR-3D to make the slowly varying field. The slowly-varying-velocities requirement also applies if the model consists of just one block (volume) with a single velocity field: The velocity may not change rapidly.
If there are large velocity contrasts in the model, we recommend that the contrast is represented by an interface, and that there is a velocity discontinuity across the interface. See the User's Guide for Version 5.0, section 6.5.4.9 for details on how to split the velocity field properly.
Example
In areas with salt bodies surrounded by sediment layers, with a quite continuous velocity increase in the sediments, the salt bodies are represented as separate blocks (volumes) with constant salt velocity, while the surrounding sediments is a single block with a continuous, smooth velocity field.
Also notice that you now have the VTI anisotropy option: You can provide the Thomsen Epsilon and Delta as constants, 2D grids or cubes individually for each block (volume) in the model. A model can be a mixture of anisotropic and isotropic blocks.
The smoothness requirements discussed above also apply to densities and the Thomsen parameters.
Receivers and shots in the modelling of PSDM traveltime tables
In the Common Shot Survey Generator you can make a 3-dimensional grid of receivers. In PSDM we use the depth points as this 3D 'receiver' grid. In practice it must be desampled compared to the actual PSDM 3D depth point grid, using for example every other or third node along each direction.
The shots in the modelling are the (regularized) shot/receiver positions on the surface.
Event attributes - what to store and how
In the Common Shot Wavefront Tracer in 'Functions>Set Event Attributes', select 'Prestack Depth Migration' to minimize the number of parameters stored for each event. You can tune the data storage further in the 'Edit Event Attribute Settings'. By default the PSDM option gives you the Static Event File Format and just one event per receiver. Increase this if you want multi-arrivals. The purpose of the sorting order ('Receiver Lines (slow-->fast)') is to fine-tune the event storage to how you will access the data later on in the PSDM process. It should not be necessary to change this parameter in initial tests.
Parameters for the Wavefront Tracing
Most default parameters for the wavefront tracing are in version 5.0 quite OK for the dimensions and level of details usually found in seismic exploration.
Set the Maximum Traveltime to match the max. traveltime recorded in the data. Notice that the max. traveltime here is for one-way propagation, from positions on the surface and down to depth points.
If high transmission angles at interfaces are involved, increase the Maximum Incidence/Departing Angle.
Use 'No Shot Similarity'. The Shot Similarity may slow down the computations in PSDM applications. The reason is that there is a very large number of receivers, thus a large proportion of the computed wave field passes through receivers. It also takes a quite long time to collect info from adjacent 'full' shots to construct the initial wavefronts for the 'intermediate' shots.
Before you start a major modelling job, model just a few shots, and check the results in the interactive common-shot tracer and with the event file extractor.
For a large job, use the batch tracer and divide into smaller sub-jobs for parallel computing.
Inspect the results for selected shots and receivers in the Common Shot Wavefront Tracer and the Event Attribute Extractor. Use the NORSAR-3D Event File toolkit to access the NORSAR-3D results from your PSDM application.