The TIGRESS group strives to provide advanced technology for evaluation by the Princeton research community. To this end, we regularly acquire small experimental systems with new technology so that research groups have an opportunity to evaluate and benchmark, to a limited extent, these advanced systems.
The experimental systems do not provide our production environment, such as /tigress or backup. They are usually allocated in one or two week units for the sole use of a research group. They are provisioned with accounts of the group members and standard software. They normally do not run job schedulers, unless they represent a fairly large resource. At the end of the allocation, all of the files on the system are discarded and a new standard environment installed.
No files on experimental systems are backed up. A lost file is gone forever.
Presently, our only experimental system is a GPU cluster called McMillan. It is described under the McMillan entry within the left menu bar. Unlike most experimental systems, it is shared by all of the Princeton research community.