Rice University high performance computing in oil and gas

PGS brings back the supercomputer for massive Triton survey. Univa trials Docker for seismics and finds its latency wanting. CGG announces GPUWrapper for AMD/Nvidia code portability.

In his keynote address to the Rice oil and gas high performance computing conference held last month in Houston, PGS’ Sverre Brandsberg-Dahl announced the return of the supercomputer in exploration seismology. For the last decade or more, clusters have performed the bulk of seismic processing, often described as ‘embarrassingly parallel.’ For PGS, this changed with the huge (10 million sq.km.) 2014 Triton Gulf of Mexico survey. This five vessel, high-fold, long offset, dual-sensor, broadband survey provides full azimuth data. To process the 660 terabyte data set, PGS built ‘Abel,’ a special purpose, massively parallel supercomputer, actually a 24 cabinet CRAY XC40. Abel is an all CPU machine (for ease of programming) and has a total 600 terabytes of memory.

Univa’s Ian Lumb looked at the potential of Docker’s container technology in oil and gas. Docker has performed well in proof of concept trials but the transition to production use requires containers to be incorporated into existing IT infrastructures and workflows. Docker users outside of oil and gas report easier replication, faster deployment and lower configuration and operating costs. But the technology is unlikely to support ‘latency intolerant’ MPI applications such as seismic imaging. On the other hand, the use of GPUs via Docker containers appears promising.

A team from CGG has been working on GpuWrapper, a cross platform API for heterogeneous GPU environments. Today’s oil and gas market is dominated by Nvidia GPUs and the proprietary Cuda programming language. The compatibility layer allows for device-independent programming across Cuda and OpenCL, and promises ‘maximum performance’ on both Nvidia and AMD GPUs from a single code base. GpuWrapper is said to bring a negligible performance hit and is ‘fully transparent’ for geophysicists. The company has over a petaflop of AMD GPUs successfully deployed in production and reports great performance on wave-equation modeling.

Bags more presentations and videos on the conference website.

Click here to comment on this article

Click here to view this article in context on a desktop

© Oil IT Journal - all rights reserved.