...
Our HPC cluster benchmarks at 10 terraflops--around 100 times the performance of a high-end workstation. It’s been designed for loads that require parallel processing of distribdistributed data sets.
Our HPC Cluster
Model | IBM nx360 M4 |
---|---|
Number of compute nodes | 48 nodes |
Node CPU | Dual Intel Xeon Processor E5-2620 v2 6C |
Total cores per node | 6 cores per CPU x 2 CPUs = 12 cores with 64GB RAM |
Hardware threads per core | 12 |
Hardware threads per node | 12 cores x 12 threads = 144 total threads |
Clock rate | 2.1GHz |
RAM | 8x 8GB (1x8GB, 2Rx8, 1.35V) PC3L-12800 CL11 ECC DDR3 1600MHz LP RDIMM |
Cache | 15MB Cache 1600MHz 80W |
Node storage | 500GB per node |
Internode network | 56gbit/second Infiniband |
Cluster storage | 108 TB of GPFS storage |
Cluster file system | GPFS / Spectrum Scale |
Operating System | Red Hat Enterprise Linux [Liam, please add version] |
...
Please work with a faculty member to sponsor your work.
Accessing the HPC Cluster
...
Shell and Data Access
Shell Access
Use SSH to access the command shell use SSH; to upload files use SFTP or SCP:
Host: login.hpc.lab
Port: 22
Credentials : your Florida Poly username and password
[Liam, I don’t understand the following.]
Submitting jobs to the LSF (use bsub)
Compile the hello_world example code provided by default
/opt/ibm/platform_mpi/bin/mpicc -o hello_world.exe /opt/ibm/platform_mpi/help/hello_world.c
Submit a job through LSF to test the message passing
bsub -n 10 -R "span[ptile=1]" -o %J.out "/opt/ibm/platform_mpi/bin/mpirun -lsf -vapi /home/(Username)/hello_world.exe; wait"
Check the output of the %J.out file to verify results
Applications
Spack
File Upload
Use SFTP or SCP to upload files with the access parameters shown above.
Applications and Packages
Libraries
MPI
The Message Passing Interface (MPI) is a library specification that allows the HPC to pass information between its various nodes and clusters.
Compiling Source
The following compilers are installed:
IBM PE Runtime (mpicc, et. al) v09.01.02.00u including C, C++, and Fortran.
GNU Compiler Collection v4.4.7 including C, C++, Objective-C, Fortran, Ada, Go, and D.
Packaged Binaries
You can install applications on our HPC using Spack: a Linux package manager that makes installing scientific software easy. With Spack, you can build a package with multiple versions, configurations, platforms, and compilers, and all of these builds can coexist on the same machine.
To list all available packages:
spack find
To load a package into your environment:
spack load
You can specify a software version as part of the load:
spack load python@3.7.3
loads Python 3.7.3
Python and PIP
If you install Python using Spack you can use PIP to install
...
other modules:
spack load python
python3 -mpip install matplotlib
Apache Hadoop 2.6.0
Apache Hadoop is a collection of open-source software utilities that facilitate using a network of many computers to solve problems involving massive amounts of data and computation. It provides a software framework for distributed storage and processing of big data using the MapReduce programming model.
...
If you need an application that’s not available through Spack please contact the Helpdesk: helpdesk@floridapoly.edu or 863.874.8888.
Submitting a Job
Linux binaries
You'll submit jobs through bsub, part of IBM’s LSF workload management platform. IBM provides documentation for bsub.
You may wish to use IBM’s mpirun script to abstract the job from underlying hardware. IBM provides documentation for mpirun.
Some examples:bsub -n 10 -R "span[ptile=1]"-o ~/hello_world.out "mpirun -lsf -vapi ~/hello_world”
bsub -n 20 -R "span[ptile=6]"-o ~/hello_world.out "python3 ~/hello_world.py”
Python
Install Python
spack load python
Install Python bsub
python3 -mpip install bsub
Run your code with bsub
bsub -n 10 -o my_job.out "python3 my_job.py"
Other
For help with other jobs please contact the Helpdesk: helpdesk@floridapoly.edu or 863.874.8888.
...