Products
Product Brochure – (VIEW) (DOWNLOAD)
Tsunami Kirchhoff Migrations
Leading-Edge Technology Developed and Optimized for Linux Clusters
The Tsunami Imaging Suite provides the ultimate capability for distributed computing in the seismic processing industry. With the ability to mix Intel, AMD, SUN and SGI nodes within the same cluster and as little as 256 mbytes of memory on each CPU, unprecedented performance can be achieved. Tsunami migrations are faster than the most popular systems with better image quality. Our exceptional scalability has been tested running over 1,000 CPU's on a single job!
The Tsunami PSTM/PSDM is a distributed computing, hardware independent, Kirchhoff implementation for 2D/3D prestack time and depth migration. Featuring numerous industry advances, the software includes automatic restart capabilities to manage compute node failures during jobs, eliminating the need for continuous job status monitoring, making projects more likely to complete on schedule.
All Tsunami products incorporate a well designed graphic user interface, developed by a processing geophysicist, with all the ease of use expected from a next generation product. It includes extensive parameter and data checking, resulting in fewer mistakes when setting up jobs, therefore fewer jobs need to be rerun.
Tsunami's Patented Data Access Method
How do we achieve exceptional performance and preserve superior image quality? Tsunami uses a patented method for delivering the data to the compute nodes within the cluster. Only a single node of the cluster, designated the master, reads the seismic traces from disk. The master node then sends the traces to a single compute node within the cluster. The compute nodes then send the trace data to each other. Using this data delivery method Tsunami can operate at full-speed using a 100 mbit network regardless of the size of the cluster and can avoid using NFS as the data delivery mechanism. Jobs have been completed with Tsunami using over 600 cpus.
Our method of data delivery reduces the load on the server and spreads the load across the network, the typical bottlenecks. This allows the user to exploit the full scalabilities and efficiencies of a large cluster. Often in other imaging applications each compute node reads the data using NFS. As the cluster grows in size, the load on the data server grows. When the number of cpus reaches about 50-60, the server cannot deliver sufficient data to keep the compute nodes busy. This forces the user to split the cluster into smaller clusters each with its own server. Other solutions can include preloading the data onto each compute node, or investing in very expensive servers and network hardware. These other methods are expensive or very time consuming. Tsunami allows the user to scale clusters to very large size, while keeping the hardware costs to a minimum.
Operating Platforms & Characteristics
• Software operates in a heterogeneous operating system environment
• Cluster nodes may be different types (AMD vs Intel; 64bit vs 32bit) different speeds, have different amounts of memory, and different numbers of cpu's
• 256 mbyte of memory per CPU for PSTM
• 512 mbyte of memory per CPU for PSDM
• Requires no disk space on compute nodes
• Only requires 100 mbit network interface
• Does not use NFS, MPI, PVM or other middleware
• Uses patented technology to distribute data from the data server to the compute nodes
• Automatically restarts job if a compute node fails
• Jobs may easily be interrupted to reassign nodes
• Licensing will allow any number or combination of jobs to run on the licensed nodes, no user restrictions
• Simple software installation on the master node only, making it easy to get started, and easy to update with new versions.
Time and Depth Migrations
• Amplitude preserving
• Adjustable three-point anti-alias filter
• Capable of imaging beyond 90 degrees
• User can define aperture as a function of time/depth
• Aperture function can vary with inline or crossline
• Sophisticated trace weighting to compensate for irregular acquisition
• Can automatically match the shape of the output area to the input area saving run time when imaging irregular land surveys
• User can adjust output frequency as a function of time/depth to improve performance.
• User can select traces to migrate based on azimuth of source/receiver locations
• Input data can be sorted in any order
• Reads FOCUS® and SeisUP® formats as well as SEGY data formats
• Reads SEGY velocity files
User Interface
The Tsunami user interface is JAVA based and is jointly developed with Interactive Network Technologies (INT). Using INT to develop the user interface allows us to rapidly add new and sophisticated functionality. The user interface guides the user through the process of setting up and starting jobs, monitoring jobs, and canceling jobs. In addition, displays of fold map, offset and azimuth histograms help the user to make decisions about the parameters used to image the dataset. The user interface can be started or stopped without affecting the imaging jobs that are running. Because of the intuitive design of the user interface new users can become productive within days.
Fold Map With Shot-Receiver Locations
 |
Left: An example of the fold map display. The background color indicates the fold, shots are yellow crosses, and receivers for a single selected shot are shown in blue. The user can turn on/off any of the layers, and zoom and pan on the map. |
Node Table Right: A Node table of the cluster environment is generated allowing the user the ability to easily select specific nodes a processing job will execute on. Note that the selected nodes can be any combination of OS, memory, speed or configuration. |
 |
 |
Tsunami GUI Left: Geometry Input Page |
Tsunami Geometry Weighting and Offset Balancing Produces Accurate Amplitudes for AVO Analysis
Tsunami uses sophisticated methods of amplitude balancing so output gathers show the proper AVO signature. Amplitudes can be distorted when traces are acquired at irregular densities, and when the population of offset bins vary by an order of magnitude or more. Using Tsunami can compensate for these irregularities so that the amplitude variation by offset is not dominated by the acquisition footprint.
| Regular Offsets, No Weighting | With Trace Offset Balance | With Geometry Weighting and Trace Offset Balance |
The data above represents highly irregular geometry taken from a South Louisiana survey merge. Input synthetic shot gathers have constant amplitude with offset. When applying trace offset balancing it creates output offset bins such that the numbers of input traces in each offset are equal. Normally offset increments are equal (i.e. 1000, 2000, 3000, etc.) and this can result in a very different number of input traces going into each offset bin. This option makes varying offset increments such that the numbers of input traces in each offset are the same.
The geometry weighting creates amplitude weights for the input traces based on the acquisition geometry of the input data. These weights are derived through an inversion process, based on the trace distribution of each offset bin. Traces that are closer together will have a smaller weight; those further apart will have a larger weight. Damping factors prevent the creation of spikes in sparsely populated areas.
www.tsunamidevelopment.com
©2005 Tsunami Development — All Rights Reserved
Tsunami Imaging Suite is a trademark of Tsunami Development.
FOCUS is a trademark of Paradigm Geotechnology, B.V.
ProMAX® is registered trademark of Landmark Graphics Corp.
SeisUP® is a registered trademark of GeoCenter, Inc.
INTViewer is a trademark of Interactive Network Technologies, Inc.
