A Short History of USP


A Funny Thing Happened on the Way to Research

This is a brief history of how scientists at the research facility of a major oil company conceived and developed a seismic processing system on unix to serve their research needs in the areas of signal analysis and wavefield imaging, how they persevered in the face of organizational politics, and how they ultimately triumphed to the benefit of both research and seismic exploration.  As you download and unpack FreeUSP please take a minute to read our history...  It sketches a journey important to us and it recognizes a number of valued colleagues along the way. In its modest way it tries to imbue FreeUSP with a little of the human element.

Like most major oil companies Amoco had its own internal proprietary processing system.  Amoco's Seismic Information System (SIS) developed over a period of many years starting in the mid-1960's when digital seismic data started to be routinely recorded in the field and when large mainframe computers became available to process those data.  The genesis for the digital seismic processing revolution however pre-dates this period by more than a decade.  It actually began with Norbert Wiener and a visionary energetic group of students at MITs Geophysical Analysis Group (GAG).  If Professor Wiener provided the seeds for revolution Sven Treitel and Enders Robinson provided abundant fertile soil for them to grow.  These two pioneers led the way laying the framework for digital data processing, including band pass filtering and the concept of the layer cake earth model and prediction error filtering.  This digital tool kit provided the foundation of every subsequent proprietary processing system.

By the early 1980's SIS had grown into a vast centralized system of some 300 individual programs serviced by an army of programmers.  Each program expected data input from digital magnetic tape and wrote data to the same medium.  The data processing center had a vast tape library and hordes of sneaker clad carbon based units to rush tapes to and from the phalanxes of tape drives.  The whole operation took up several floors of a large office building and in Amoco's case was duplicated at several locations around the world. As SIS grew and as exploration success became more dependent on data processing modifying programs or adding new ones based on research became more onerous.  Custodians of SIS mutated into guardians - almost a priesthood -  entrenched in conservative practices.  Changes in programs required traversing a bureaucratic labyrinth with a mean travel path measured in months.  A year or more might elapse before a new program found its way through the system of coding specs, coding, testing, documentation, and certification.  This led to a data processing system that was very stable but incredibly lethargic to change.  Portability of this system and its cousins in other exploration companies was another huge problem since  much of the code tended to be specific to one vendor's hardware - good for the vendor, bad for the user.  For Amoco porting SIS to the Perkin-Elmer machines, then with better price/performance, was a project of epic proportions, very little of which actually advanced the state of the art of the tool kit but certainly provided employment for legions of twinkie consuming professional programmers.

During this time fundamental research into digital signal analysis and wavefield imaging continued.  The Robinsons and Treitels in turn produced a new crop of students eager to turn their ideas into practical results.  At Amoco's Tulsa research facility a number of these scientists gradually grew frustrated with the disconnection between research and production data processing: trying new algorithms on amounts of seismic data large enough to be considered part of the exploration process. By the mid 1980's this desire to wed signal analysis research and seismic interpretation became paramount. At that time happy circumstance provided one of Amoco's first unix boxes, an HP9000 with 3mb of memory, three cpu's, and 400mb of disk.  The machine was part of a purchase to provide the core of a system to process digital acoustic log data in real time in the field.  It quickly became apparent that the unix operating system was an ideal framework upon which to build a more general data processing system, one that was disk oriented relying on the unix structures of stdin, stdout, and stderr to handle data flow from one independent process to another.  This contrasted with several vendor systems of the time which were essentially giant fortran wrapper programs which built a data processing flow by linking desired object code into a single runtime program.  The premise was that only the fortran (and later the C) compiler could be relied upon to be  portable across hardware.  Such systems tend to be very finely tuned making contributions by a loosely knit coterie of scientists with disparate coding skills problematic.  As the  unix tide of popularity surged across a sea of hardware vendors USP was born in the summer of 1985.

The original gang of four were Don Wagner, Paul Gutowski, and Martin Smith, all Amoco research people, and Prof Bob Herrmann of Saint Louis University, then acting as a consultant.  Within a year the system comprised about 50 programs and was capable of processing most 2D seismic data sets from field tapes to final interpretable stack.  USP quickly demonstrated that scientists who were not primarily computer programmers could rapidly implement ideas and test them on significant quantities of data.  Also it became clear that such a system, relying upon unix for its infrastructure, did not require people intensive maintenance and in fact could be kept alive and growing by just a few cone headed scientists. The system was then adopted by other research groups within the Amoco Tulsa geoscience community.  During the late 1980's several varieties of unix boxes became available - Sun, Cray, DEC, IBM - and so considerable thought went into design of a makefile system that would support network connection of multiple architectures.  Leading this effort were Martin Smith, Terri Fischer, and Joe Wade.  As USP continued to develop it was aided by the efforts of Amoco's seismic imaging research group -  MBS - which contributed many of the tools used to migrate seismic data.  Leading scientists at the time were Dan Whitmore, MaryAnn Thornton,  Sam Gray, and Gary Murphy.  At this point some adventurous and rebellious geoscientists in Amoco's office in Cairo saw the potential for addressing severe noise problems in data recorded in the Gulf of Suez by a system that so intimately combined research, data processing, and interpretation.  Paul Garossino, Mike Bush, and Dennis Yanchak developed a highly effective methodology for reducing obscuring noise that was within weeks implemented within USP and quickly made available in production mode.  This development had a major impact on exploration success in this area.  It also had a major impact on Paul's career who became so enamored of the USP concept that he abandoned his exploration career for one in research joining the Seismic Signal Analysis group permanently in 1990.

The early 1990's saw troubled times for USP which for all its technical success and acceptance by research scientists ran afoul of Amoco's corporate IT managers, who saw it as a variance in their efforts to implement a one size fits all standard in data processing.  The efforts to kill USP lasted several years and were not dissimilar to efforts almost two decades earlier of AT&T managers attempting to kill unix itself.  The USP crew even proposed giving the system away in some public fashion but Amoco management at the time refused.  The illogic of trying to kill USP on the one hand while refusing to part with it on the other provided an interesting philosophical backdrop to the war weary scientists trying to do research.  What ultimately saved USP within Amoco was a growing user community in the exploration offices who saw their jobs made more effective by the powerful data analysis capabilities of the system combined with the very rapid turn around of new ideas into functioning tools.  This trend was facilitated by a training frenzy over several years during which more than 100 Amoco geoscientists attended a four day USP class taught by Garossino and Gutowski.  Another major boost to USP came from Amoco's New Orleans office which in 1993 led by technical geophysicist Mark Truxillo decided to exclusively use USP as its internal processing system, quickly demonstrating its technical effectiveness and economic impact.  At about that time Amoco's strategic applications technology group led by Craig Cooper saw opportunities presented by USP for plugging into research much earlier in the development cycle and enhancing his group's vision of staying ahead of the technology curve thus giving Amoco's exploration efforts competitive advantage.  By the mid 1990's USP included a fully functional 3D processing capability built around Amoco's legacy core 3D system.  It also included some multicomponent processing tools.

By the mid to late 1990's USP was firmly established within Amoco as the system of choice for deploying new technology.  During this time the USP crew included scientists James Gridley and Greg Partyka and computer analysts Marilyn Miller and Jacquie Vinson, as well as Paul Gutowski, Paul Garossino, Don Wagner, and Joe Wade.  USP now included sophisticated seismic attribute technology and cool tools to access data directly from interpretive workstations such as GeoQuest, Landmark, and Charisma.  Now an interpreter could graphically build a USP flow utilizing many cpu's over the network and stream his data through it depositing the results directly into his workstation for subsequent interpretation.  Supported architectures included Suns, HPs, Cray/UNICOS, Convex, IBM/AIX, SGI, DEC, and PC/Linux. The tool kit now had some 400+ programs and was capable of processing any digital data, including gravity, EM, and medical images.

When BP and Amoco merged in 1999 a strategic decision was made to retain USP and a core group of scientists familiar with its use. BP scientists quickly became users and developers and the future of USP became more of a sure thing.  Our post merger crew now includes Paul Garossino, Paul Gutowski, Don Wagner, Ganyuan Xia, Greg Partyka (all signal analysis, petrophysics, or imaging specialists of various vintages) and Joe Wade and Bryan Helvey, our systems gurus. In 2000 the USP guys again raised the issue of some kind of public release of USP.  The motivation was to enhance BPAmoco's technical reputation, to provide scientists and students in academia with a fully functional development and processing system, and in the spirit of open source distributions aggregate the contributions of other scientists into the system.  This request was met with enthusiastic support of management and initiated a year long effort on the part of the USP crew (in addition to their regular work load) to ready the system for release as part of FreeUSP.  What you the downloader have in your hot little hands is the fruit of many scientists' research efforts spanning almost half a century from the dawn of the digital revolution.  We thought it also important to tell you something of the story of USP.

BACK Return To About Us.