National University of Singapore

Release of Atlas6 and LSF queue updates

Dear HPC Users,

We are pleased to announce the release of atlas6, a new HPC cluster which joins the pool of HPC compute resources at Computer Centre. The cluster has 47 compute nodes (and a login node) with each node having dual hexa-core Intel X5650 2.66GHz processors and 48 GB of memory (eight of the nodes have 96 GB of memory). Thus, the new cluster has a total of 6.1 TFLOPS computing power and 2,688 GB of memory.

New queues are created for atlas6 cluster and some changes are made to the queues for the atlas5 cluster to cater for different requirements and optimizing the system usage.

Please note the following:

1. Logging in

The login node (for ssh) is atlas6_c01. You may also use the name atlas6 which is aliased to atlas6-c01. To access atlas6 batch resources and work space, users could also login to atlas5. Atlas6 may also be accessed via HPC Portal (Run Jobs / Apps).

2. LSF batch queues (atlas5, atlas6)

On atlas5 cluster, the original queues 8core_parallel, 4core_parallel and a5openmp will be closed and phased out after all jobs have completed.

Atlas5 and atlas6 have large memory and high number of cores and are especially suited for large serial, OpenMP (shared memory) and parallel jobs. These clusters also join the atlas3 and atlas4 clusters in servicing the serial queue linux64. Please use the following queues on atlas5 and atlas6.

Job description	Queue on atlas5	Queue on atlas6
Serial large memory	atlas5_large (up to 20GB)	atlas6_large (up to 72GB)
Parallel	atlas5_parallel (up to 24 cores)	atlas6_parallel (up to 36 cores)
Test	atlas5_parallel_test	atlas6_parallel_test
OpenMP	openmp	openmp
Serial	linux64	linux64

 

3. Where to submit your jobs (work space)

Atlas6 shares with atlas5 the parallel GPFS file system located at /atlas5tmp. As there is a limited bandwidth between the home file system (/home) and the atlas5/atlas6 clusters, please submit large and parallel jobs from /atlas5tmp.

> cd /atlas5tmp
> mkdir my_subdir
> cd my_subdir
> bsub …. ./my_executable

More details on the high performance work space may be found here. 

4. Compiling your code

Due to differences in the Intel processors, executables compiled on atlas3 and atlas4 cluster may not run on atlas6. To ensure compatibility, please recompile your code on atlas6-c01. The following compilers are available on atlas6:

Compiler	Command
INTEL	ifort, icc
MPI	mpif90, mpif77, mpicxx, mpicc
GNU	gfortran, g77, gcc

More details on the compilers and code compilation may be found here.

Please check the HPC website for updated information and instructions. For problems or queries, please contact us at .