Spectrum Spatial (SSA/LIM)

?We are looking at improving SSA performance and investigating allocating dedicated GPU resources.

Are there any SSA Flags / command line options available that can be included to offload Java processing to a GPU if available?

Have the SSA Java Jar files been compiled to take advantage of GPU Processing?

?Thanks Amita, sorry to say the link comes back with an invalid page for me.  ..I did notice a message appeared briefly that mentioned chat or chatter not being enabled.

cheers

Thanks that's great,

I look forward to hearing anything else in regards to the SSA Java GPU? options.

cheers

SSA uses Spectrum Spatial for a lot of its functionality: (e.g. tiles and data access). Neither Spectrum nor Spectrum Spatial has any special code to take advantage of any GPU that may be available. I suspect the same is true for SSA with the possible exception of the javascript running in the browser. Some browsers offload graphics operations to the GPU.

What are you trying to do with the GPU?

Essentially we are optimising our infrastructure and taking advantage of MS Server 2016? latest functionality, which enables provision of dedicated GPU resources. In this case allocation to VDI is the best fit given that Spectrum cannot leverage this. Any plans to develop with GPU FPU offloaded processing in mind?

Sounds like you just want Spectrum Spatial and Analyst to be as fast as possible no matter whether it uses VDI or not. Although a semi-active area of interest and research there are no plans to add capabilities to Spectrum Spatial to leverage VDI at this time.

Thanks for this great info, ?

 Hi Michael,

Essentially to speed tile processing server side, offloading javavm processing via jcuda comes to mind.

Infrastructure wise, we can now allocate the full 4GB GPU to a single virtual server running within a hyperV host in our cluster. If this was supported by Spectrum Spatial this is definitely where the greatest advantage would be.

While client side allocation of GPU across the VDI user desktops speeds up internet browser rendering for all websites. This doesn’t scale well as GPU resource is shared and becomes an issue for failover of desktops. There needs to be enough unallocated free GPU resource on each HyperV host in the cluster in case of failover. So not as great an advantage but still useful.

Thanks again for your feedback, it's useful to know for future infrastructure planning.

Cheers