TJ Receives Decommissioned NASA Servers
Thanks to TJ’s Mentorship Program, the Computer Systems (CS) Lab has a $300,000 IBM server rack, courtesy of NASA. Last year, CS Lab Director Dr. Shane Torbert (pictured with the rack) and Mentorship Director Alfred Lampazzi visited NASA’s Center for Climate Simulation (NCCS) to meet with the mentor of a senior in the CS Lab. During the visit, Dr. Torbert learned that when NCCS replaces its servers, as it does periodically, it makes an effort to see that decommissioned equipment is reutilized rather than processed through equipment disposal, especially if the equipment is only several years old.
Once NCCS has decommissioned a piece of equipment and determined that no other NASA center or project has an interest in it, the agency notifies other non-profit organizations of the equipment’s availability. Pursuant to the Stevenson-Wydler Technology Innovation Act of 1980, all US educational institutions are eligible to receive excess research equipment from NASA for use in technical and scientific education and research activities.
Having expressed an interest in an IBM server rack that had been used for climate modeling, Dr. Torbert filled out the required paperwork. Upon receiving the rack this October, he explained why it was such a meaningful acquisition:
“The 40 nodes each contain two Sandy Bridge CPUs, each capable of eight parallel cores, for a total of 640 simultaneous threads. There are 2 GB of RAM per core for a total of just over 1 TB of memory. Interconnect speeds between the nodes are over Mellanox Technologies InfiniBand, which is capable of data transfer at a rate of 40 gigabits per second, 40 times faster than the lab’s 1 gigabit internal network and 400 times faster than common 100 megabit networks. Speeds this fast make possible the use of Remote Direct Memory Access. Each node also contains a Xeon Phi, Intel’s alternative to the NVidia CUDA GPU used throughout the lab.
“This acquisition builds upon a previous donation from Northrop Grumman of a Voltaire (now Mellanox) model Infiniband system.”
TJ students will use the servers for parallel computing and for projects requiring cloud-type virtualization, such as Hadoop, which currently run off third-party systems.