16 | Halliburton Landmark
of faults in seismic data compared with older methods and ignores many of the non-faultrelated structures . Additionally , the scanning step generates strike and dip information for the highlighted discontinuities ( Figure 1F ), which can be used in many workflows , such as for the later fault plane extraction step .
Despite the superior results , two factors have limited the adoption of the fault likelihood attribute . The first is the large number and complexity of the parameters . Many test runs are typically required for every dataset to optimize the output . The second , and more critical , problem is the sheer computing power required to generate fault likelihood volumes accurately .
CLOUD COMPUTATION TO THE RESCUE
To generate traditional fault attributes , one or a few calculations are performed per sample in a dataset . To generate the fault likelihood attribute , potentially hundreds of calculations must be run per sample depending on the range of strike and dip orientations selected . This makes it very challenging to run fault likelihood on traditional workstations for all but the smallest datasets . To overcome this problem , Halliburton Landmark developed a cloud-native implementation of fault likelihood , available as part of Seismic Engine . Seismic Engine , a DecisionSpace 365 cloud application , provides a comprehensive suite of cloud-native seismic attributes . It parallelizes and distributes these complex algorithms , allowing you to leverage the scalability of the cloud . This scalability enables you to use whatever
Figure 3 > A comparison of workstation performance with the performance of Seismic Engine , a DecisionSpace ® 365 cloud application . The columns represent how long each method takes to generate a fault likelihood attribute for various seismic data sizes .