NCAR Machine Room
I started working at the National Center for Atmospheric Research as a student assistant in 2004. I came to Boulder for my Ph.D. because I wanted to work with supercomputers at NCAR, and four years later I was hired to be a full-time NCAR Scientist specializing in high-performance computer systems.
For my first four years at NCAR, I didn't get into the machine room very much. But as the Technology Development Division became more involved in running computer systems for research (e.g., NCAR's TeraGrid systems), I become a regular visitor to the machine room, often going downstairs just to take care of things myself. Some days I write academic papers; other days I pull cables.
The first series of pictures here were taken as part of a tour that Tom Engel put on for visiting prospective Ph.D. students in 2004. The latter pictures were taken by myself and my colleagues as we were working in the machine room.
Entering the Machine Room
The hallway to the left shows the equipment door to the machine room. As part of tours, Tom Engel usually describes NCAR's first Cray supercomputer, which is still on display in a public cooridor in NCAR's Mesa Lab facility.
In addition to providing computational resources, NCAR's Computational and Information Systems laboratory maintains several petabytes of data on spinning disk and tape-based archival systems. Interestingly, NCAR keeps several operational antique storage devices, such as spooled tape systems, to assist with data retrieval when old spools are found in attics. In 2004, many tapes were kept on shelves and required manual mounting by an operator; upgrades to the StorageTek silos increased the silo capacity leaving offline tapes for backup purposes. In 2010, NCAR's Powderhorn StorageTek silos were retired and replaced with newer variants; here I am inside a silo helping load tapes.
Bluesky - 2004
This is the first real supercomputer I was ever allowed to touch: NCAR's 1600 processor IBM SP4, named Bluesky. Bluesky was the 6th computer on the Top500 list when I arrived in Boulder, but as the supercomputing field heated up again, it quickly dropped to 9th, 11th, 15th, and then down to the 40s before it was replaced.
Facilities Tour - July 2009
In 2009, Michael Oberg convinced Aaron Anderson to give a tour of the machine room and the infrastructure support facilities to the SIParCS interns. This is when I discovered that NCAR has Cutler-Hammer breakers just like featured in Jurassic Park.
Frost Expansion - August 2009
In August 2009, NCAR acquired three Blue Gene/L racks from the San Diego Supercomputer Center (SDSC). These three racks were from SDSC's "Intimidata" system. When they decomissioned the system, we integrated these racks into our existing one-rack Blue Gene/L system "Frost", resulting in an operational four-rack system. While IBM assisted with wiring the Blue Gene/L torus and performing the system integration, we handled most of the expansion cyberinfrastructure, including Ethernet and software, within our group.