aerospace
Using goal-driven optimisation in Ansys simulation software, the geometry of a UAS wing was modifi ed to achieve the desired improvement in lift. The result: a slightly modifi ed platform that can deliver a much heavier payload
Platform LSF intelligently schedules
parallel and serial workloads while using an organisation’s available computing resources at maximum capacity and capability. Like many engineering fi rms, MTU works its clusters around the clock. Every CPU is tapped for as much compute power as can possibly be extracted. Engineers submit jobs from commercial applications like ABAQUS and
MSC.Nastran as well as from several in- house codes, the most important of which is a fl uid dynamics application, which accounts for around 90 per cent of cluster usage. Axel Philipp, MTU’s Platform LSF
Load scheduling for
commercial and in-house codes
Savings can also arise when HPC systems are put to use in the best way. Consider an example from MTU Aero Engines, a large German manufacturer of commercial and military engines, which has a large Linux cluster in Germany and a smaller one at a subsidiary in North America. Around 150 users submit an average of 3,000 jobs a day to this HPC environment. The company maintains a mix of different size hosts (memory, number and type of CPUs), as some of the applications used on the cluster cannot take advantage of the newer multi- core CPUs, because of memory bandwidth issues, while others request large amounts of memory. Platform LSF from Platform Computing has a number of features that help work with this heterogeneous environment.
www.scientific-computing.com
administrator, estimates that the utilisation rate of the 1,400 CPUs in the main cluster averages at around 80 per cent over a typical year, while some months see usage exceeding 90 per cent. ‘With Platform LSF we can achieve high utilisation while still getting free slots for those who really need them,’ states Philipp. ‘That’s what I need that software for –
hardware capacity without these scheduling capabilities, and this translates into 150 fewer cores. Delaying the cost of purchasing additional blades is benefi cial in itself, but it also saves the cost of powering and cooling those extra CPUs.
Optimising a standard product for customer requirements Now let’s shift attention to those vital components working in the background. Shaw Aero Devices, a part of Parker Aerospace, manufactures valves for aerospace applications, and a customer needed fuel control valves for an unmanned aerial vehicle. Six valves operate in a manifold to distribute fuel among tanks and keep the aircraft in balance as fuel is consumed. The vehicle’s manufacturer wanted to increase the payload by using a smaller fuel pump. However, the customer specifi ed a pressure drop of
WITH PLATFORM LSF WE CAN ACHIEVE HIGH UTILISATION WHILE
STILL GETTING FREE SLOTS FOR THOSE WHO REALLY NEED THEM. THAT’S WHAT I NEED THAT SOFTWARE FOR – FAIRSHARE SCHEDULING TO PRIORITISE JOBS AND PRE-EMPT LESS IMPORTANT ONES
fairshare scheduling to prioritise jobs and pre- empt less important ones.’ Separate queues are also set up for in-house and commercial licenses. Philipp explains that commercial licenses get higher priority, because ‘they’re very expensive, so we need to make sure that they’re being used when needed.’ Philipp estimates that MTU Aero Engines would need at least 10 per cent more
0.75psi at a fl ow rate of 4.45gpm while Shaw’s standard valve specifi ed 6.09psi. Rob Preble, project engineer for Shaw,
created the original design in AutoCAD and simulated it using FloEFD computational fl uid dynamics software from Mentor Graphics. FloEFD’s use of native 3D CAD data, automatic gridding of the fl ow space and managing the fl ow parameters as
OCTOBER/NOVEMBER 2011 45
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52