http://www.sgi.com/International/UK/stories.dir/proudman_ocean.html (Einblicke ins Internet, 10/1995)
The general trend in scientific computing over the last few years has been to move from a shared mainframe resource to individual desktop workstations. As these workstations have decreased in size and price, their capabilities have increased enormously, allowing scientists to analyse their data more quickly and thoroughly than ever before. One scientific institute, the Proudman Oceanographic Laboratory (POL), makes particularly sophisticated demands on its network of workstations. POL is one of the marine science institutes within the Natural Environment Research Council (NERC). The laboratory has been involved in the development of computer simulations for investigating the motion of the sea since the 1960s. Much of the impetus for this work has been the need to establish the correct height of sea defences to protect against tidal surges. Early work included a model of the Thames and the southern North Sea from which the design parameters for the Thames barrier were established. These days people look to POL to provide the analysis of such events as the flooding of Towyn on the North Wales' coast in 1990 and north Norfolk in 1993.
POL also operates the British Oceanographic Data Centre (BODC) which was founded to maintain and safeguard a national oceanographic database. Understanding and predicting the ocean's behaviour requires the collection and processing of enormous numbers of measurements covering a wide diversity of variables and involving many disciplines. Not only is there an enormous richness of interacting physical, chemical and biological phenomena to investigate, but modern-day oceanographic research often calls for the coordination of many teams of researchers and technologists with each team targeted at specific areas.
"Some areas of science are more expensive to research than others.
Oceanography is on a par with nuclear physics and astronomy as one of the
most expensive areas for information gathering. Ocean-going survey ships
cost upwards of 10,000 per day, and these need to be equipped with much
specialised instrumentation, which all adds to the cost. It is therefore
essential that we optimise the analysis of the data collected, and produce
results as quickly as possible," said Steve Loch, a systems coordinator at
POL.
There are two further factors that are at the heart of oceanographic research but which add to the difficulty of oceanographic investigation. Firstly, the fundamental characteristic of the sea is that it is constantly in motion. This motion is in response to a combination of factors, most notably the tidal influence of the moon and the sun, the force of the wind, and the differences in density of the water that can arise from place to place. Time is therefore an essential element of any scientific description of the sea. Measurements taken today are not equivalent to those taken yesterday.
Secondly, the sea is clearly three-dimensional. As scientists tend to think in terms of pictures or diagrams rather than words, the difficulties of visualisation have represented and continue to represent a significant barrier to research: much progress to date has been on the basis of two-dimensional approximations. Consequently, workstations with 3D graphics capabilities open up new possibilities for gaining some insight into the way water masses move.
The North Sea Project
POL hosted the recently completed North Sea project which involved the collaboration of some 200 scientists from a dozen laboratories. The aim of the North Sea project was to establish a computerised water-quality model of the southern North Sea, which can be used to investigate man's impact on the marine environment, including the fate of nutrients, heavy metals and other pollutants introduced through the river systems. A model of this size and complexity currently requires supercomputing power, with 25 gigabytes of data output for every model month. The initial stages of the research involved a period of data collection requiring some 15 months of dedicated ship time. The ship, the RRS Challenger, was used to survey a track of 2,000 nautical miles every lunar month. The temperature and salinity of the water were measured along the track, together with approximately 20 other variables relating to its chemistry and biology. A typical cruise returned 30,000 such measurements, totalling 600,000 values. Challenger was also used to deploy and recover instrument packages such as current meters from fixed locations in the sea. These returned yet more data.
"Examining this amount of data requires access to powerful visualisation tools. Problems with the data had to be identified quickly, before the next month's cruise," said Loch.
BODC undertook the management of the data for the North Sea project, including the production of the project CD-ROM, using software developed previously in connection with its archiving role. Data archiving
Since marine data is so expensive to collect, and the amount collected to date is minute when compared with the vastness of the world's oceans, it make sense to ensure the continued availability of that data which has already been collected. BODC is the designated repository for most data collected by UK oceanographers.
"As scientists move on and change jobs, valuable datasets can be lost. No set of conditions is ever repeated exactly, so that every body of data is unique, and a reference centre is essential," said Loch.
Compiling this archive, or data banking, requires the ability to inspect the data visually for possible problems. It has to be done quickly as there are no more than two or three people within BODC to confront the output of hundreds of scientists.
Until 1986, hardcopy plotting caused a severe bottleneck in the archiving process. The modellers also faced similar problems when confronted with the increased output of their models, whose sophistication and predictive power had been steadily increasing. Waiting for information to be plotted out on paper in a separate room could sometimes mean hours of waiting.
3D Graphics Workstations
To meet the increasingly demanding graphics needs of both modellers and data archivers POL acquired its first two workstations, along with a new mainframe, in 1987. The workstations were Silicon Graphics 2400Ts, and according to Loch, no other vendor could offer comparable graphics performance. At 40,000 each, these workstations represented a considerable investment for POL but one which has proved extremely valuable. Six years later the mainframe has been replaced by a 500-gigabyte Mass Store, and a network of 53 Unix workstations, including 41 IRIS Indigos and six Personal IRIS(tm) computers from Silicon Graphics.
"Silicon Graphics has proved to be extremely competitive on price,
graphics performance, computational performance, and delivery. Each Indigo
can outperform the old mainframe in running numerical models, and the
Oracle database, hosted by the Personal IRIS 4D35, was benchmarked at six
times the speed of the mainframe. All this computing power means that it
is possible to investigate or model aspects of the sea in ways that were
previously impossible," said Loch.
On the modelling side, the workstation can produce in under three seconds a frame, that would have needed 40 sheets of paper on the plotter. Scientists can create a new view of a 3D scene as required, virtually instantaneously. They can create real-time images in graduated colour, with curved surfaces, and lighting & shading where appropriate.
"Within BODC we have received data from over 130 organizations in a comparable number of different formats. Handling this diversity of data with just three or four people necessitates a generalised approach and considerable software expertise. Software development and debugging is now much easier in a multi-windowed environment and we are no longer constrained to work with large datasets on tape as in the days of the mainframe," said Loch.
BODC's need for the rapid appraisal of data has been met by the in-house development of the SERPLO (SERies PLOtting) program. This is a 12,000-line interactive Fortran program utilising Silicon Graphics' Graphics Library (tm) GL(tm). Rather than having to examine column after column of raw data or printed graphics, scientists can now see any erroneous values at a glance on the screen and edit them out where appropriate, thus achieving enormous savings in time and effort.
BODC has also recently completed the first phase of EDTEVA, a SERPLO derivative, which is used to display and edit tidal data from the UK's Class `A' network of some 40 tide gauges. It is anticipated that staff effort will be reduced to a tenth of the previous level.The streamlining of the work at BODC is vital, since the organisation has to service a wide user community. Over the past six years it has responded to more than 1000 requests from scientists and engineers in 175 different organisations at home and abroad.
The ability to visualise the complicated 3D structure of the sea and how it changes with time is becoming increasing critical to advances in oceanography. Silicon Graphics workstations are making a huge difference to research at Proudman Oceanographic. The high processing power and sophisticated 3D graphics on each desktop allows scientists to reduce considerably the time interval between data collection and presentation of the results, making it much easier for them to justify the investment made in their research.
The work at BODC has also been facilitated by the use of Iris Indigos for archiving, ensuring that important scientific data does not gather dust in corners, but instead is easily accessible.
