====== COR@L Lab ======

COR@L Lab is the computational software system in Lehigh University Industrial and Systems Engineering Department. 

===== Hardware =====

^ Sl. No.  ^ Name           ^ Memory (MB)  ^ CPU                             ^ Scratch  ^ Operating System        ^ Remarks                                                      ^
| 01       | coral          | 4000         | Intel Xeon 3.10GHz (4 cores)    | --       | Debian 10      | webserver, mailman, bugzilla, mysql, trac, subversion. 64 bit  |
| 02       | polyp1 (polyps)  | 32000      | AMD Opteron 2.0 GHz (16 cores)  | 761GB    | Debian 12      | 64 bit architecture, head node / login node                         |
| 03       | polyps (2-15)  | 32000        | AMD Opteron 2.0 GHz (16 cores)  | 761GB    | Debian 12      | 64 bit architecture, head node: polyp1   |
| 04       | polyps (30)    | 128000       | Intel Xeon 2.4GHz (24 cores)    | 92GB     | Debian 12      | 64 bit architecture, Tesla K80 GPGPU, head node: polyp1        |
| 05       | beluga         | 11000        | Intel Xeon 3.2GHz (4 cores)     | --       | Debian 8       | 64 bit architecture, storage node, exports home, software, and globalscratch directories        |
| 06       | shark          | 2000         | Intel Xeon 2.4GHz (2 cores)     | 250GB    | Ubuntu 14.04   | Software licences.                                             |

===== Software =====

====== Polyp30 ======
Access to polyp30 is accessible via PBS. Please use an interactive job to (qsub -I) to gain access to its resources. Once you have a job, you can submit your batch job to that node. It has Nvidia Tesla K80 GPGPU. Currently it has in /usr/local the following versions of cuda installed:

<code>user@polyp30:~$ cd /usr/local
user@polyp30:/usr/local$ ls -al |grep cuda
drwxr-sr-x 17 root   staff 4.0K Jun 26  2016 cuda-7.5
drwxr-sr-x  8 root   staff 4.0K Oct 10  2017 cuda-8.0
drwxr-sr-x 19 root   staff 4.0K Mar 26  2018 cuda-9.1
drwxr-sr-x 18 root   staff 4.0K Mar 27  2018 cuda-9.0
drwxr-sr-x 18 root   staff 4.0K Dec 16  2024 cuda-10.2
lrwxrwxrwx  1 root   staff   21 Jul 22 08:06 cuda -> /usr/local/cuda-11.4/
drwxr-sr-x 16 root   staff 4.0K Jul 22 08:08 cuda-11.4</code>

Please update your PATH (echo $PATH) and any other variables to utilize the version you need. (LD_LIBRARY_PATH, PYTHONPATH).

===== K80 GPGPU Support =====
nVidia ended Cuda support for the K80 GPGPU with Cuda 11.4. Newer versions will not function with this GPGPU. 

[[ https://developer.nvidia.com/cuda-legacy-gpus|nVidia Legacy GPGPU matrix]]

<code>Support for the following compute capabilities are deprecated in the CUDA 11.0 Toolkit:
    sm_35 (Kepler)
    ► sm_37 (Kepler) ◄
    sm_50 (Maxwell)

See matrix link above
</code>

* [[https://docs.nvidia.com/cuda/archive/11.0/cuda-toolkit-release-notes|CUDA Toolkit v11.0.3 Release Notes]]

* [[https://docs.nvidia.com/cuda/archive/11.4.4/cuda-toolkit-release-notes/index.html|CUDA Toolkit v11.4.4 Release Notes]]

* [[https://docs.nvidia.com/cuda/archive/11.4.4/cuda-toolkit-release-notes/index.html|Tesla K80 and discontinued driver support for Kepler]]

  
===== Hints =====
1. Due to the heavy usage of the login node (polyps / polyp1), it is recommended you launch an interactive job on a compute node. This will provide you the dedicated resources you request rather than running on a shared node.

===== Administrators =====

  * Cor@l admin mail-list: coral-admin-list [at] lehigh.edu

  * Mark Motsko (mjm519 [at] lehigh.edu)
  * Suresh Bolusani (bsuresh [at] lehigh.edu)