March 25, 2014

Making HPC and Big Data More Accessible

Tiffany Trader
SGI UV cropped

Dating back to 2009, the SGI Altix UV line has always targeted performance and scalability. In this recent video, SGI addresses how this inclusive HPC and big data computing environment can help meet the needs of the larger scientist and analyst user population. It’s a platform that’s innovative yet stays within standards, according to Dr. Eng Lim Goh, CTO of SGI.

The SGI UV platform currently scales up from 4 to 256 Intel Xeon sockets and 64 TB of memory running on a single off-the-shelf Linux OS. With the NUMAlink interconnect, SGI UV 2000 enables scaling a single system starting as small as 16 threads up to a maximum of 4,096 threads.

“What’s most important about this server is it runs standard Linux,” observes Goh.

To test the compatibility of Linux, Goh purchased an off-the-shelf NVIDIA card. He loaded VNC on a server remotely and and called it up from his laptop. To further test the compatibility, he loaded LibreOffice and the R statistical package, and plugged in the RGL renderer. He demonstrates the output visualization.

Goh also wanted to show how data analysts could access this extreme scale system from their mobile devices. To do this the SGI engineers took an iPad and installed the iTeleport software from the official App Store to run VNC. Then you simply select the server you want to visualize and the same thing that is on the 256-socket server will be displayed on the laptop and on the mobile device, says Goh.

To further test the limits of mobility for the coming era of wearable devices, Goh’s team beamed the display onto the Google Glass, allowing users to visualize their outputs while moving about.

Goh concludes the presentation by summarizing three primary features of the SGI UV system.

1. It scales up from four sockets all the way to 256 sockets.
2. It scales to 64 terabytes of memory.
3. It runs standard Linux, so that it will be compatible with Linux software.