Tesla cluster in SERC is consist of three compute nodes. Each compute node is an SMP node built using 16 AMD-Opteron cores housed in 4-Quad-core CPUs. Each of these compute nodes is also connected to a NVIDIA-Tesla S1070 GPGPU node. Each Tesla node is composed of 4 GPUs with each GPU made up of 240 processor cores. The cluster is managed by PBSPro workload manager to distinguish and allow compute as well as GPU based jobs. Each compute job can use a maximum of 14 CPUs on this cluster since multi-node jobs are disabled. For GPU-based jobs each GPU needs a CPU-bound thread to drive the computation on it and hence the compute node CPU-resources are divided into two PBS virtual nodes namely the cpu-node and gpu-node. The jobs that get to be scheduled on these virtual nodes are identified based on the PBS job script variables as described under the workload manager section. The user needs to define appropriate PBS variables to define whether his jobs are GPU based or only CPU-based. Based on these variables PBSPro workload manager automatically routes the job into execution queues to schedule to appropriate vnodes. Each GPU is configured to be used in exclusive mode by a job and the job can use one or a maximum of 4 GPUs at a time. The compute jobs can use MPI or OpenMP based codes and the GPU jobs are built using the NVIDIA CUDA libraries.
Hardware Overview :-
System Softwares/Libraries :-
Application Softwares/Libraries :-
Portable Batch System Professional (version 10.0)
Recent Activities on CUDA Programming
Location of Tesla Cluster
Hostname of the machine
Accessing the systemThe Tesla cluster has one login node,tesla1, through which the user can access the cluster and submit jobs. The machine is accessible for login using ssh from inside IISc network (ssh email@example.com). The machine can be accessed after applying for basic HPC access, for which:
For any queries email to firstname.lastname@example.org or please contact SysAdmin #103 SERC.