The Microsoft-sponsored survey by Gulf Publishing unit Gulf Research* we report on on page 8 found that ‘Microsoft operating systems dominate high performance computing (HPC) in upstream oil and gas.’ More precisely, 96% of the respondents reported HPC use of Microsoft Windows, ‘on a daily basis,’ while 73% of the sample reported that they ‘never used’ Linux for HPC.
When I read this, I thought I had woken up from a dream to find myself in a parallel universe! My impression, having tracked this issue for some time (see our Technology Watch report from the 2000 Microsoft in HPC meet, now in the public domain on www.oilit.com), was that Microsoft was in the position of a ‘wannabe’ trying to gain ground from Linux’ huge success in this space.
Only last December, we reported from the High Performance Computing Session at the SEG (Oil ITJ Vol. 11 N° 12) where IBM HPC guru Earl Dodd who stated that ‘Commodity Linux clusters have repealed Moore’s Law and now dominate the HPC landscape.’ Dodd noted extremely high take up in oil and gas, the second HPC vertical after government.
Reading as I do, press releases on the results of surveys on this, that and the other, one is inured to the bold forecasts made for, say the take-up of RFID technology over the next 10 years, the monetary benefits of full deployment of ‘digital oilfield’ technology or even how many engineers the oil and gas business will employ a few years down the road. Generally, such burning issues are addressed by ‘researchers’ phoning round some folks who a) have a job to do and b) don’t know the answer to any of the above questions. The results are predictably random. You might as well ask people to give their opinions on successive digits in the expansion of Pi, although it might be harder to monetize the results.
But the Microsoft survey is not a forecast, but ostensibly a survey of the status quo. I had to check my facts. When the old brain is called into question, I turn to Wikipedia, font of a lot of, if not all, knowledge and particularly well informed in IT matters. What does Wikipedia have to say about HPC? I quote, ‘The term high performance computing (HPC) refers to the use of (parallel) supercomputers and computer clusters, that is, computing systems comprised of multiple processors linked together in a single system with commercially available interconnects. [...] Because of their flexibility, power, and relatively low cost, HPC systems increasingly dominate the world of supercomputing. Usually, computer systems in or above the teraflop-region are counted as HPC-computers.’
Well there’s a clue to the mystery. If you phoned around folks engaged in upstream IT, how many would have a Teraflop at their disposal? How many would know what a Teraflop was even? But I digress. Returning to my Wikipedia oracle, I asked, ‘where do I go to learn more about HPC and what operating systems are used?’ The response was, ‘The most powerful high performance computers can be found on the Top 500 list—www.top500.org.’
Hey, if it’s that easy, I thought, why didn’t Microsoft just publish the URL of the Top 500 and save all that money on the ‘Survey.’ The answer could be that the Top 500 list paints a very a different picture of HPC. The latest survey, carried out in November 2006 (the list is updated every six months) confirms Linux’ domination of HPC. Approximately 75% of the machines on the TOP500 list are running one flavor or another of Linux. The only other systems that rate anywhere are IBM’s AIX and HP-UX with 8% and 5% respectively. Microsoft does not even make it onto the list!
Compute Cluster Server 2003
This got me wondering if Microsoft’s definition of HPC was somehow different from the TOP500 list. So I went onto Microsoft.com and did a search for ‘HPC’. As I expected, this brought up a list of references to Windows Compute Cluster Server 2003. So HPC for Microsoft is about clustered COTS** computing—just like the TOP500.
Oil and gas
The TOP500 list is self reported and probably does a poor job of counting industrial deployment of HPC. Many companies consider the details of their HPC deployment as commercially sensitive and keep the number to themselves. Could oil and gas be different from the Top 500? Certainly not seismic processing, which has spearheaded the move from the large ‘shared memory’ architectures of the past to clustered COTS architectures running Linux. Reservoir engineering has seen a less spectacular paradigm shift. Flow modeling is less ‘embarrassingly parallel’ than seismic processing.
Compute Cluster Server 2003
Microsoft could have solved the mystery by instructing its researchers to ask the simple question, ‘How much of your computing is performed on Windows Compute Cluster Server 2003 (CCS)? With the ‘march of the Petrel’ into reservoir modeling, and the enthusiasm for Microsoft’s solutions demonstrated by some major oil and gas accounts, there must be good reasons for departmental level HPC leveraging Windows CCS. It would have been more credible for Microsoft to have reported even modest numbers on such possible deployments, rather than the marketing delirium of the ‘Survey.’
** Common off-the-shelf i.e. commodity hardware—a large number of interconnected PCs.
© Oil IT Journal - all rights reserved.