Working with Rice’s Research and Cost Accounting (RCA), IT redesigned the SPICE (Shared Pool of Integrated Computing Environments) charging model to fit requirements established by grant funding organizations such as the National Institutes of Health (NIH) and the National Science Foundation (NSF). Although Rice researchers receive a designated amount of storage space for their projects, researchers in fields requiring big data and high performance computing clusters often require additional space, or they may need a virtual machine (VM) in which to run their code. The new charge model approved by RCA included storage increments called bricks. A VM Brick consists of either one, two or 4 cores and 40 Gigabytes (GB) of disc space, reserved for the operating system. A Storage Brick consists of one Terabyte (TB) of storage.
The storage space and VMs purchased under the new SPICE model are applied to solutions already bought by researchers, so there are no additional charges, only a revised charging model. SPICE appeases the researchers who need any number of TBs of space (minimum of one) and/or a research computing space separate from their desktop computer, but do not need the more vast resources for High Performance Computing (HPC) provided through the Research Computing Support Group (RCSG). However, it should be noted that the VMs in this model are meant to handle moderate to low computations.
“SPICE, a Shared Pool of Integrated Computing Environments, represents an evolutionary step in providing centrally-managed storage and servers to Rice researchers. It represents a model to achieve our “condo computing” model into the server and workstation based research arena. SPICE provides to Rice University’s Researchers a cost and time saving ability.” said HPC Support Supervisor, Joseph Ghobrial.
On July 28, 2014, we migrated the STIC scheduler/resource manager from MOAB/TORQUE to the open-source solution SLURM (Simple Linux Utility for Resource Management (http://slurm.schedmd.com/)).
While MAUI/TORQUE works well for uncomplicated queuing requirements like those of SUGAR and BlueBioU, SLURM is required for more complicated condo environments like STIC and DAVinCI.
SLURM was developed by a group at the Lawrence Livermore Nation Laboratory(LLNL) and is now required in all large scale lab procurements which forces system vendors to get behind SLURM in order to bid on national lab procurements. Many of our academic colleagues (Harvard, Duke, Oxford, Michigan, Buffalo, etc.) have already moved to SLURM.
For more information about SPICE, visit: http://www.rcsg.rice.edu/spice/
For more information about the SLURM migration, visit: http://www.rcsg.rice.edu/stic-migrating-from-torquemoab-to-slurm/