Tamela Bolen

Frontera, the world’s largest educational supercomputer housed on the Texas Superior Computing Middle (TACC), is massive each by way of number of computational nodes and the capabilities of the massive Memory Wave "fat" compute nodes. A few recent use circumstances exhibit how educational researchers are utilizing the quad-socket, 112-core, 2.1 TB persistent memory to assist Frontera’s massive Memory Wave App nodes to advance a large number of analysis matters together with visualization and filesystems. The arrival of Software Outlined Visualization (SDVis) is a seismic occasion in the visualization community as a result of it permits interactive, high-resolution, photorealistic visualization of giant information with out having to move the info off the compute nodes. In transit and in situ visualization are two methods that permit SDVis libraries resembling Embree and OSPRay to render data on the same nodes that generate the information. In situ visualization renders data for visualization on the identical computational nodes that perform the simulation.

In transit visualization lets customers tailor the render vs simulation workload by utilizing a subset of the computation nodes for rendering. "The HPC group is getting into a new era in photorealistic, interactive visualization utilizing SDVis," stated Dr. Paul Navrátil, Memory Wave App director of visualization at TACC. The quad socket Intel Xeon Platinum 8280M large memory Frontera nodes give scientists the flexibility to interactively render and see necessary events (because of CPU-based mostly rendering) and - again interactively - bounce back in the data to study what precipitated the vital occasion to occur. This interactive "instant replay" capability is enabled by the high core rely, high-bandwidth (six memory channels per socket or 24 memory channels whole) of the TACC massive memory 2.1 TB nodes. Jim Jeffers (senior principal engineer and senior director of superior rendering and visualization at Intel) has been a central mover and shaker in HPC visualization together with his work on SDVis and the Intel Embree and Intel OSPRay libraries.

He explains, "Optane Persistent Memory offers scientists with the memory capacity, bandwidth, and persistence options to allow a brand new level of control and capability to interactively visualize large knowledge units in real time and with up to film-high quality fidelity. Scientists are in a position to acknowledge or more simply identify key occurrences and interactively step ahead and backward in time to see and understand the scientific importance. David DeMarle (Intel laptop graphics software engineer) factors out that the 2.1 TB memory capacity within the Frontera large memory nodes offers customers the ability to keep intensive histories of their OpenFOAM simulations in memory. Using software, scientists can set off on an event, obtain an alert that the event has happened, and then overview the causes of the event. Collisions, defined as an occasion the place a number of particles are contained in a voxel or 3D block in space, are one instance of an vital fluid circulate occasion. Alternatives embody triggers that occur when the stress exceeds or drops below a threshold in a voxel.

Memory capability is essential to preserving the simulation histories that assist scientists perceive physical phenomena as trendy systems can simulate bigger, more complex programs with increased fidelity. Protecting data in the persistent memory units delivers a performance boost. DeMarle observes, "The runtime savings is extremely correlated to quantity of memory, which implies that the savings will scale to massive runs each by way of size and resolution." Scalable approaches are necessary as we move into the exascale computing period. DeMarle and his collaborators utilized in situ methods to create their OpenFOAM visualizations and histories so the information doesn't have to maneuver off the computational nodes. They called the Catalyst library to perform the in situ rendering. Alternatively, customers also can carry out in situ visualization utilizing the OpenFOAM Catalyst adapter. ParaView was used as the visualization instrument. To manage resource utilization, Catalyst calls the open-supply Intel memkind library. This gives two advantages: (1) the persistent memory capability could possibly be allocated for use by the simulation (utilizing Memory Mode) and (2) data could possibly be immediately written to the persistent memory gadgets using App Direct mode.