Biocluster is a high performance computing (HPC) cluster, and it is the oldest member of the Wang lab. It was ordered and built even before the Wang lab officially started at the University of Southern California (USC), around July 2010. From 2010 to 2016, Biocluster was located in the basement of the Zilkha Neurogenetic Institute, Keck School of Medicine of USC. In September 2016, Biocluster was moved to be hosted at the Department of Systems Biology Information Technology (DSBIT) group at Columbia University. The photo was taken when Biocluster was just moved to the new home, in many server boxes.
Over the years, a lot of hardware and software changes have been made to biocluster, to improve its performance and increase its storage capacity. Currently, Biocluster serves almost all the computing needs for our laboratory, and is a highly trusted friend for all lab members.
A tutorial on administrating biocluster is available here.
Currently, Biocluster has about 50 computing nodes; most of them have 24 Intel X5675 @ 3.07GHz cores and 96Gb memory, four of them have 64 Intel E5-2683 v4 @ 2.10GHz cores and 256Gb memory, one of them have 32 E5-4627 v2 @ 3.30GHz cores and 1Tb memory, and one of them have ~30,000 GPU cores in 8 Titan X GPUs. All nodes are connected by 40Gbps infiniband connections.
Additionally, Biocluster has about 1.5 petabyte of storage in four separate storage servers, which serves different needs of the Wang lab. This include one super-high performance scratch server composed of enterprise SSD drives organized in RAID 0 configuration, as well as three storage servers using enterprise hard drives in RAID 6 configuration.
Biocluster uses Rocks cluster management system, and uses Sun Grid Engine as the job scheduling system.