EAGE workshop on open source software in geosciences

Academics revel in open source software for (mostly) geophysics. dGB shows how a ‘freemium’ business model can work. But oils are wary and seem to prefer proprietary consortia.

Introducing the 2016 EAGE workshop on open source software in geoscience, physics Filippo Broggini (ETH Zurich) observed that open source software (OSS) is ‘taking over,’ on internet browsers, on mobile phones (with Android) and especially in high performance computing. The ensuing Workshop showed that while there is enthusiasm for OSS in academia, its penetration into the upstream is so far rather limited.

dGB Earth Sciences’ Kristofer Tingdahl was the only presenter to address the issue of creating a viable business around OSS. Users are interested in vendor independence and object to the lock-in and high maintenance costs of commercial software. While dGB’s Open dTect seismic interpretation package is open source, there is little code coming back from users who are not in general developers. dGB is working to expand its community of developers with a Python interface. The trick, for dGB and others, is to strike a balance between OSS purism and commerciality. Tingdahl’s dream is one of an open cloud infrastructure supporting a DDS-based (see below) storage layer with hooks for JavaSeis and Madagascar for seismic processing, Open dTect for interpreting and other OSS software for geology and mapping.

Sergey Fomel (U TX Austin) traced ten years of Madagascar’s development. The OSS seismic processing package was rolled on at the June 2006 EAGE. The package exceeded expectations and is now in release 2.0, with 40,000 downloads and 260 ‘reproducible’ papers published. Madagascar is comparable to Matlab but with better handling of array data on disk. Madagascar benefits from OSS-style governance, all contributors have equal control, there is no decision tree hierarchy. The ‘CircleCI’ continuous integration platform is used to perform thousands of tests prior to a new commit. Fomel concluded with a pointer to a new ‘Geophysics papers of the future’ initiative to kick off in 2017.

John Stockwell (Colorado School of Mines) presented OpenSeaSeis, again with a Matlab analogy. The package allows sophomores to process seismics and provides a data viewer and QC package. The package is well written but ‘incomplete.’ It provides trace by trace processing and interactive workflows. 3D is under development.

Jeffrey Schragge (University of Western Australia) has deployed Madagascar in the cloud, specifically on Australia’s Nectar research cloud. Nectar offers ‘SME* computing’ to smaller processing shops and universities without big clusters. Again Madagascar (a.k.a. M8R) is available along with the ‘Scons’ python interface that recognizes parallel parts of code. A ‘challenging’ 3D full wavefield model of the Australian NW shelf required 100 million core compute hours! ‘Burst cloud computing’ and dynamic pricing made this possible. But Schragge warned of the risk of bidding wars for compute resources. CYthon, HDF5, ZeroMQ, NumPY also ran. Cuda accelerators are also an option. The toolset is being bundled as an M8R extension ‘mycloud.py.’

In the concluding discussion, the perennial issue of researching and reading the plethoric seismic data formats was raised with a suggestion for an industry-backed open source API. For John Stockwell, this already exists, ‘it’s called DDS.’ While we need to avoid creating new data formats, conversion is a problem that might be amenable to a common API. It was suggested that the academic community might ‘lead the charge’ here as industry was unlikely to get involved. HDF5 was suggested as an ‘ideal universal format’ as was Apache Arrow, an open source format for big data. The requirement (and difficulty) of involving the SEG Standards committee was also an issue. Another problem for the OSS community is the availability of good real world (not Marmousi) data. There is need for an OSS data repository. In fact, this is in the process of being created as a component of the SEG Wiki. Finally, the issue of how upstream research is funded was raised. OSS code is a poor fit with the sponsored consortia paradigm. Consortia are ‘sold’ to oils by researchers and OSS ‘is not a good business reason.’

Comment - While the OSS geoscience movement exists (in geophysics at least), it is mostly constrained to academia and even there, suffers from a degree of fragmentations as long term projects proceed in parallel. The meeting concluded with a suggestion that another gathering could be organized ‘in five years.’ This movement proceeds at a snail’s pace. But the SEG open seismic processing workshop (August 17-20, Houston) is another step in the right direction.

* Small and medium enterprise.

Click here to comment on this article

Click here to view this article in context on a desktop

© Oil IT Journal - all rights reserved.