Sie befinden sich im Service: RWTH High Performance Computing (Linux)

Login Nodes

Login Nodes

Kurzinformation

    Login nodes are login servers within the HPC Cluster that users can use to connect and access all computing resources.

    To connect, please note that:

    • Multi Factor Authentication (MFA) is mandatory to connect to any login node. A detailed explanation on how to configure MFA can be found here.
    • A connection to login nodes is only possible from within the network of the RWTH, most German universities, as well as some research institutions. Details on how to setup and use a VPN connection to the RWTH network can be found here. If you want your German university or research institution to be added to the trusted networks, get in contact with us.

    Use these login nodes to connect to the HPC System:

    Login NodeSpecial FeatureConnection Type
    Claix-2023
    login23-1.hpc.itc.rwth-aachen.de-SSH, Console
    login23-2.hpc.itc.rwth-aachen.de-SSH, Console
    login23-3.hpc.itc.rwth-aachen.de-SSH, Console
    login23-4.hpc.itc.rwth-aachen.de-SSH, Console
    login23-x-1.hpc.itc.rwth-aachen.dex-server runningRemote Desktop
    login23-x-2.hpc.itc.rwth-aachen.dex-server runningRemote Desktop
    login23-g-1.hpc.itc.rwth-aachen.de

    with H100 (MIG)

    (comming soon)

    SSH, Console
    Claix-2018
    login18-1.hpc.itc.rwth-aachen.de-SSH, Console
    login18-2.hpc.itc.rwth-aachen.de-SSH, Console
    login18-3.hpc.itc.rwth-aachen.de-SSH, Console
    login18-4.hpc.itc.rwth-aachen.de-SSH, Console
    login18-x-1.hpc.itc.rwth-aachen.dex-server runningRemote Desktop
    login18-x-2.hpc.itc.rwth-aachen.dex-server runningRemote Desktop
    login18-g-1.hpc.itc.rwth-aachen.deNvidia GPUSSH, Console
    login18-g-2.hpc.itc.rwth-aachen.deNvidia GPUSSH, Console
    Jupyter
    https://jupyterhub.hpc.itc.rwth-aachen.de:9651CPU + NVIDIA GPUBrowser, Jupyter

    Please Note: For maintenance reasons all login nodes are rebooted each Monday morning at 6 am.

    These login nodes can be used for programming, compiling, debugging, preparation and postprocessing of batch jobs.

    They are not intended for long and intensive computation runs of programs! Processes that have consumed more than 20 minutes CPU time may get killed without warning. 


    IMPORTANT:

    • Computing on the login nodes is discouraged. Please use the compute nodes!
    • Compiling, coding and result viewing can be done on login nodes, but will be faster on a compute node!
    • Using too many resources for any kind of task on a login node will trigger automated and manual responses from us, resulting in your work being lost!
    • This limitation is actively enforced. We reserve the right to terminate any process on the login nodes to maintain a good quality of service for all other users.
    • You may also receive a verbal and written warning!

    Detailinformation

    Technical Details

    This limits apply only to the login-nodes, unless specified otherwise.

    If you need a graphical session without these limits, please submit an interactive batchjob with X11 forwarding to the compute nodes.

    If you need a console session without these limits, please submit an normal salloc job to the compute nodes.

    CPU

    We use cgroups on the login nodes to configure all user processes and enforce limits. 

    Within a login node, we limit all processes of a user to a maximum of four (4) cores. We may also kill processes that use more than 20 minutes of CPU time, especially when a system becomes overloaded.

    Real physical memory (RAM)

    We limit the total amount of physical memory  (+swap) to 25% (of the total) that any user can use across all of their running processes. The following command displays the current memory quota usage:

    memquota

    Unfortunately, if a user hits this memory limit, the login-node may be unresponsive for a few seconds while the limits are enforced. As all users are affected by this, it is considerate and take this memory limits into account from the beginning.

    Virtual memory

    We do not enforce the amount of virtual memory you can use or allocate, but if your programs start  reaching to physical memory, the physical memory limit will be enforced!

    Network and Disk

    We enforce network and disk limits manually to guarantee a good quality of service for other users.

    Please use one of the dedicated systems copy18-1, copy18-2, copy23-1, or copy23-2 for applications that transfer data or need high network bandwidth, e.g. copy processes.

    zuletzt geändert am 05.04.2024

    Wie hat Ihnen dieser Inhalt geholfen?

    Creative Commons Lizenzvertrag
    Dieses Werk ist lizenziert unter einer Creative Commons Namensnennung - Weitergabe unter gleichen Bedingungen 3.0 Deutschland Lizenz