What is Gremlin?

Gremlin is the name of a visualization cluster computer system at the Mississippi State University HPC². A visualization cluster is a computer system created by assembling a collection of workstation systems into a single computing entity, and to combine their graphics capabilities in order to mimic the visualization capabilities of much more expensive graphics server systems.

We began research into clusters in 1987 and have written a brief history of our clustering research.

How big is it?

Gremlin is comprised of a total of twenty rackmounted desktop workstations housed in two racks. The workstations are configured as follows:

• 4 IBM Intellistation Z Pro Workstations (super nodes)
  • 3.06GHz Intel Xeon processor
  • 2 gigabytes of RAM
  • 80 gigabytes of disk space (dual 40GB IDE drives)
  • 5 network interfaces (4 gigabit ethernet, 1 100Mb/s ethernet)
  • NVIDIA Quadro FX3000G graphics card with 256MB of video memory
  
  
• 16 IBM Intellistation M Pro Workstations (render nodes)
  • 3.2GHz Intel Pentium IV processor
  • 2 gigabytes of RAM
  • 80 gigabytes of disk space (dual 40GB SATA drives)
  • 2 network interfaces (1 gigabit ethernet, 1 100Mb/s ethernet)
  • NVIDIA GeForce FX5950 graphics card with 256MB of video memory

Gremlin uses the following networking components for inter-node communications.

• 1 Extreme Networks 48si 100Mb/s ethernet switch with gigabit uplink for external communications.
• All inter-node communications is handled via a switchless gigabit ethernet network using a collection of dedicated links and crossover cables. This diagram provides more detailed information about how this is configured.

The DVI interfaces on the four super nodes are attached to a single IBM T221 22.2" LCD flat panel monitor. Each node drives one quarter of the display at a resolution of 1920x1200. The total resolution of the monitor is 3840x2400 (QUXGA-W) for a total of 9,216,000 pixels at 204 pixels per inch.

Installed software includes the RedHat Linux operating system, MPI/Pro, MPICH, and LAM/MPI for parallel communications. DMX (Distributed Multihead X) is used to create a single X Windows session across all four displays and Chromium is used for distributed OpenGL.

What is it used for?

Gremlin is used for a wide variety of visualization tasks including climate, weather, and ocean modelling, remote sensing, and computational fluid dynamics.

What does it look like?

We have a gallery of photos available as well as a diagram that shows how the ethernet and visualization networks are configured.