ALOMA – A Framework for Seismic Applications

Complex computations on ever-growing amounts of data are characteristic challenges for seismic data processing. ALOMA is a failure tolerant runtime system that helps dealing with these challenges by executing workflows on large-scale distributed systems. As survey areas are getting larger and larger, computer systems have to adapt as well to process larger amounts of data. Using distributed resources in an efficient way is therefore crucial.

Changing hardware systems, however, are forcing geophysicists to learn about HPC techniques in order to efficiently run their software on large-scale systems. Thanks to ALOMA, they can again focus on their area of expertise and don't have to bother with parallelization, multi-threading, and other challenges in high-performance computing. Our software system takes care of the efficient execution of the algorithms – even on large scale and heterogeneous systems.

Seismic and HPC Experts Developed Together

The fundamental idea behind ALOMA is to free the geoscientists from having to learn about HPC tools and strategies in order to execute their software in a scalable way. Our software bridges the knowledge gap between geophysics and HPC experts.

Computer- and geo-scientists together came up with ideal strategies for parallelization, data partition, and failure tolerance in the context of geophysical applications. The heart of ALOMA, its failure tolerant runtime system to execute workflows on distributed systems, was then developed by HPC experts.

For the geophysicists and geologists ALOMA is merely a black box in which they can integrate their latest developments through a well-defined interface. New algorithms and prototypes can be added and tested to production scale within no time. Existing codes and applications – even in different programming languages such as C/C++, Fortran, Matlab etc. – can be integrated as modules in ALOMA.

Exemplary workflow with ALOMA
© Fraunhofer ITWM

Exemplary workflow with ALOMA: Two-dimensional input data is corrected and then summed up to one-dimensional data. Input and results can be viewed with any program for visualization of seismic data.

The Main Features:

  • Plugin architecture for easy integration of existing algorithms in various programming languages
  • Use of existing legacy binaries without the need for any sources
  • Test newly developed algorithms on production scale problems 
  • Graphical interface to create even the most complex workflows in a user friendly way
  • Automatic generation of parameter GUIs with an expandable Parameter Description Language
  • Parallel runtime environment for seismic applications that parallelizes across cores and sockets
  • Automatic adaption to data dependencies in between modules of a workflow
  • Automatic resource management and tools for execution monitoring
  • Failure tolerant execution with automatic re-scheduling of failed tasks, and automatic adjustment to addition or loss of resources
  • Existing visualization tools can be attached for instant graphic inspection of the results

ALOMA is a specialized version of GPI-Space which is widely used in fields beyond geophysics such as big data and machine learning. We were able to prove the feasibility of this concept in various projects with partners in the oil and gas industry, where we managed to make customer software scale within a few days.