HPC user accounts
An HPC User account provides access to all three clusters maintained by the NYU HPC team. If you don't have an account, you may apply for an HPC user account.
Old HPC clusters
Prince is the new HPC cluster that is currently being deployed. Prince will replace the HPC Mercer Cluster.
Mercer has 4 login and 394 compute nodes:
Number of nodes CPU type and speed Number of cores per node GPUs per node Total memory per node Memory available to jobs Node names Node set name 112 Intel Xeon E-2690v2 (Ivy Bridge) x86_64 3.0GHz (2014) 20 64GB 62GB compute-14-* to compute-20-*
48 Ivy Bridge x86_64 3.0GHz (2014) 20 192GB 189GB compute-21-* to compute-23-*
68 Westmere x86_64 2.67GHz (2010) 12 24GB 23GB compute-4-* to compute-8-7
8 Westmere x86_64 2.67GHz (2010) 12 48GB 46GB compute-8-8 to compute-8-15
16 Westmere x86_64 2.67GHz (2010) 12 96GB 93GB compute-9-*
64 Westmere x86_64 3.07GHz (2011) 12 48GB 46GB compute-12-* and compute-13-*
64 Nehalem x86_64 2.67GHz (2009) 8 24GB 23GB compute-0-* to compute-3-*
1 Nehalem x86_64 2.27GHz (2009) 16 256GB 250GB compute-10-0
1 Westmere x86_64 2.67GHz (2011) 32 1TB 1000GB compute-10-1
4 Westmere x86_64 2.67GHz (2011) 12 1 x NVidia Tesla M7020 24GB 23GB compute-11-*
9 Sandy Bridge x86_64 2.0GHz (2014) 16 4 x NVidia Titan 128GB 126GB gpu-23-*
To restrict a job to a specific subset of nodes, you can request the node set name as a feature, eg:
#PBS -l feature=ivybridge_20p_64GB_3000MHz
You can see a map of nodes and usage with pbstop. The diagram below indicates which nodes belong to which of the above nodesets:
Dumbo is a 44 data node Hadoop cluster running Cloudera Distribution of Hadoop (CDH).
- For a detailed description of dumbo and how to access it, please see the dumbo wiki pages.