ParaView on JURECA

​JURECA provides

• 1872 compute nodes and

• 12 visualization nodes equipped with
  • Two NVIDIA K40 GPUs per node
  • 2 x 12 GiB GDDR5 memory
  • 10 nodes with 512 GiB memory
  • 2 nodes with 1024 GiB memory

• Login nodes providing direct user access

There are several ways of how to do Remote Visualization with ParaView on JURECA:

Remote Visualization		pro	contra
X Forwarding (not recommended)		easy to use (ssh -X ...)	sending X/OpenGL commands is slow X Server on user workstation required (not a problem on Linux, might be a problem on Windows and Mac) X Server has to be compatible with OpenGL many OpenGL extensions need direct access to GPU
VNC See sample VNC ParaView session		easy to use, user only needs to install VNC viewer (not the vis app itself) very convenient, full desktop access of remote machine image delivery is fast GPU not needed on remote machine	software OpenGL not integrated in all VNC Servers software OpenGL is slow no parallel visualization
Client-Server mode on visualization node(s) See sample session		hardware accelerated OpenGL is fast rendering on GPU ParaView server can be run in parallel(but number of vis nodes limited)	user has to install ParaView on his workstation batch job needed ssh tunnel needed
Client-Server mode on compute node(s) (without X/GPU) See sample session		ParaView server can be run in parallel on a really huge number of nodes in situ visualization possible	user has to install ParaView on his workstation ssh tunnel needed

---

any feedback welcomed - h.zilken@…, goebbert@…