Welcome to the HPC @IMG @UNIVIE and please follow these steps to become a productive member of our department and make good use of the computer resources.
Welcome to the HPC @IMG @UNIVIE and please follow these steps to become a productive member of our department and make good use of the computer resources.
The Jet-cluster serves a [jupyterhub](https://jupyterhub.readthedocs.io/en/stable/) with a [jupyterlab](https://jupyterlab.readthedocs.io/en/stable/) that launches on the jet-cluster compute nodes and allows users to work directly on the cluster as well as submit jobs.
Goto: [](https://jet01.img.univie.ac.at) from within the VPN or UNI-Network.
Login with your jet-Credentials, choose a job and the jupyterlab will be launched.
There are several kernels available and some help can be found:
-[Python/](Python/)
-[Tutorial on Jet](Python/Your-First-Notebook-onJet_v2.ipynb)
## User Quotas and Restrictions
Memory limit / slurm requests
:construction:
### Network drives
Currently there are two network drives mounted on the jet cluster. These are connected to the [SRVX8](SRVX8.md) and data can be transfered or accessed like this on the jet cluster. Be aware that the data needs to be transfered via the network before and latencies are higher.
```
131.130.157.8:/raid61 400T 378T 22T 95% /raid61
131.130.157.8:/raid60 319T 309T 11T 97% /raid60
jetfs 1.1P 81T 975T 8% /jetfs
```
# System Information
Last Update: 4.12.2020
Last Update: 4.12.2020
Node Setup
Node Setup
- 2x Login Nodes
- 2x Login Nodes
...
@@ -55,8 +27,6 @@ Node Setup
...
@@ -55,8 +27,6 @@ Node Setup
## Node Information
| Name | Value |
| Name | Value |
| --- | --- |
| --- | --- |
| Product | ThinkSystem SR650 |
| Product | ThinkSystem SR650 |
...
@@ -69,7 +39,7 @@ Node Setup
...
@@ -69,7 +39,7 @@ Node Setup
Global file system (GPFS) is present on all nodes with about 1 PB (~1000 TB) of storage.
Global file system (GPFS) is present on all nodes with about 1 PB (~1000 TB) of storage.
# Software
## Software
The typcial installation of a intel-cluster has the INTEL Compiler suite (`intel-parallel-studio`) and the open source GNU Compilers installed. Based on these two different compilers (`intel`, `gnu`), there are usually two version of each scientific software.
The typcial installation of a intel-cluster has the INTEL Compiler suite (`intel-parallel-studio`) and the open source GNU Compilers installed. Based on these two different compilers (`intel`, `gnu`), there are usually two version of each scientific software.
Major Libraries:
Major Libraries:
...
@@ -122,5 +92,68 @@ nco/4.9.3-gcc-8.3.1-g7o6lao
...
@@ -122,5 +92,68 @@ nco/4.9.3-gcc-8.3.1-g7o6lao
on how to use environment modules go to [Using Environment Modules](Misc/Environment-Modules.md)
on how to use environment modules go to [Using Environment Modules](Misc/Environment-Modules.md)
The Jet-cluster serves a [jupyterhub](https://jupyterhub.readthedocs.io/en/stable/) with a [jupyterlab](https://jupyterlab.readthedocs.io/en/stable/) that launches on the jet-cluster compute nodes and allows users to work directly on the cluster as well as submit jobs.
Goto: [](https://jet01.img.univie.ac.at) from within the VPN or UNI-Network.
Login with your jet-Credentials, choose a job and the jupyterlab will be launched.
There are several kernels available and some help can be found:
-[Python/](Python/)
-[Tutorial on Jet](Python/Your-First-Notebook-onJet_v2.ipynb)
## User Quotas and Restrictions
Please try to use the compute nodes in a responsible manner. Currently there are not restrictions on the duration or the resources you can request. However, please follow these rules of collaboration:
- Memory, keyword: `mem` e.g. each Node up to 754 GB
- Runtime, keyword: `time` e.g. try to split jobs into pieces.
Consider the following example. You can use one node relatively easy for more than 3 days with your jobs running, but do not use all nodes an block them for all other users for 3 days. If you need multiple nodes, split the jobs into shorter runtimes. In general it is better to have more smaller jobs that are processed in a chain. Also try not to use too much resources that get wasted. Have a look at resources used in your jobs using the `/usr/bin/time` command or look [here](https://gitlab.phaidra.org/imgw/slurm).
Sample Job
```
#!/bin/bash
# SLURM specific commands
#SBATCH --job-name=test-run
#SBATCH --output=test-run.log
#SBATCH --ntasks=1
#SBATCH --mem=1MB
#SBATCH --time=05:00
#SBATCH --mail-type=BEGIN # first have to state the type of event to occur
#SBATCH --mail-user=<email@address.at> # and then your email address
# Your Code below here
module load miniconda3
# Execute the miniconda Python
# use /usr/bin/time -v [program]
# gives statistics on the resources the program uses
# nice for testing
/usr/bin/time -v python3 -v
```
Storage Limitations are set mainly to the HOME directory (default: 100 GB), but there are some general restrictions as well.
On the Login Nodes (jet01/jet02) jobs can only use 20 GB of Memory as the rest needs to be reserved for the file system and destribution services. Jobs will be automatically killed after a safety margin.
## Network drives
Currently there are two network drives mounted on the jet cluster. These are connected to the [SRVX8](SRVX8.md) and data can be transfered or accessed like this on the jet cluster. Be aware that the data needs to be transfered via the network before and latencies are higher.
These scripts are intended to help with certain known problems.
Report problems to: michael.blaschek@univie.ac.at
```
These are scripts in a common directory. Feel free to copy or edit as you like. Note that some services like filesender require an ACONET account (accessible via your u:account).
These scripts are intended to help with certain known problems.
Report problems to: michael.blaschek@univie.ac.at
```
These are scripts in a common directory. Feel free to copy or edit as you like. Note that some services like filesender require an ACONET account (accessible via your u:account). Please note the available VNC services.
## Virtual Machine Hub
## Virtual Machine Hub
Currently the system acts as a virtual machine host.
Currently the system acts as a virtual machine host.
