High-Performance Computing 2019-20

HPC Yearbook 19/20

Tis would require a direct link to the

satellite sources and constant access to HPC resources. Böhnke said: ‘Working with live satellite data is one of the main issues in this project. As cost-intensive fixed assets, we have to work within the boundary the satellites were originally designed for. In the same way, we aim to generate strong recommendations into the design of future sensor technologies.’

Speedy visual

Visualisation is a core element of the VESTEC project to enable fast and accurate decision making. However, workstations at the crisis management centres are usually not able to process and render the huge volumes of data produced by the supercomputer. Tus, the project needs to move the extraction and rendering of visualisation objects to the HPC resources. During the simulation computation,

the most relevant visualisation features are extracted and streamed to the decision- maker, which considerably reduces the data stream to the crisis management centre. Gerndt added: ‘Tis will lead to real-time decision making with a precision not possible before.’ Te project is using the ParaView

solution as one method to visualise the data from these simulations and the Catalyst solution as the mechanism for connecting the simulation codes to ParaView. ParaView, combined with Catalyst, allows

The monolithic approach of high-performance computing has to be broken up to support real-time data streams from sensor sources like satellites, surveillance cameras, or even social media networks

as Brown added: ‘Previous approaches were able to rely on special arrangements with HPC machine owners to prioritise jobs, or even kick out running jobs and replace them. However, in our approach, this is not sufficient, because we could have very many jobs running over a long period of time but with a workload that, at any specific moment, is fairly unpredictable.’ Instead, the Vestec system tracks the

status of each HPC machine and, when a job must be run, a decision is made about the most appropriate supercomputer to execute it, which is entirely seamless from the user’s perspective. Tis is not without challenges, as every HPC machine is different. Te project also requires interactive “

live visualisation of the simulation running on the HPC, which brings interactivity to the end-user. Tis allows the user to see and analyse (using post-processing capabilities of ParaView) the result of the simulation immediately, without having to wait until the end of it. Michael Migliore, an R&D engineer on

supercomputing for high-velocity data stream processing, and needs to establish real- time visualisation streams. However, most supercomputing centres usually do not allow such interactive access to their resources, and the team is working around this lack of accessibility. Te hardware also needs to allow for

immediate results on big datasets, linking HPC machines to live sensor data. Christian Böhnke from the German Remote Sensing Data Center at DLR, explained: ‘What we are longing for is a system retrieving new data whenever [it is] available. Tis may cover time intervals of 15 minutes or one hour, so the HPC machine must listen to a data server or constantly request new data.’

the scientific visualisation team at Kitware, said: ‘ParaView Catalyst is the glue between the HPC architecture, the simulation code and other libraries. Improving this ecosystem with better integrations of the libraries and better Catalyst adapters in the simulation code would be major steps.’ Intel also contributed its ray tracing-

based visualisation technology to bring the simulations to life. Miroslaw Pawlowski, a graphics soſtware engineer from Intel, said: ‘Ray tracing is a great technique, not only for movies but also for scientific visualisation, because it scales very well with dataset size (hundreds of GBs) and its ability to directly support different geometric objects (without tessellation into triangles), in combination with volume rendering.’


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