1 of 20

Jupyter  Iceland Project

Sigurður Detlef Jónsson

Implementation for the Icelandic Research-E Infastructure IREI 

IREI

2 of 20

Agenda

  • Introduce UTS and Myself
  • Usecases
  • Unicore
  •  Unity IdM
  • Architecture
  • Plan of Implementation

2/20

2024

3 of 20

About Me

3/20

2024

  • Currently working in the IT department at University of Iceland

    • working on the IREI project as a system administrator

  • Education:

    • BSc. degree In Software Engineering from Reykjavík University

    • Finishing MSc. degree In Software Engineering at University of Iceland

4 of 20

IREI Infrastructure

OBJECTIVE

Develop a powerful infrastructure to support and service research in Iceland

Which include:

IT consultancy,

access to powerful HPC,

solutions for data storage  

and data sharing

GRANTS

Infrastructure fund has granted 455 m.kr to the project so far

Foundation for the infrastructure is up and running

Further development and implementation underway

USAGE

Over 230 registered users, a significant increase from 80 2 years ago 

4 IREI specialists 

Knowledgebase and experience growing 

Number of users rapidly increasing

4/20

2024

5 of 20

IREI Services

RÁÐGJÖF

Í boði er … 

Almenn UT ráðgjöf

Aðstoð við að finna lausnir

Innleiðing lausna

Umsjón með innkaupum

MIÐLUN GAGNA

Örugg miðlun, innan lands sem utan

Umhverfi fyrir samstarf

GAGNAGEYMSLA

Örugg og skilvirk varðvarsla rannsóknargagna

Upplýsingaöryggi og persónuvernd

Hagkvæmni – greind í stjórnun varðveislu

REIKNIAFL

HPC (CPU and GPU)

Gagnageymslur vegan reikniaðgerða

Mikið reikniafl fyrir hraða úrvinnslu rannsóknargagna

CONSULTANCY

We offer: 

General IT consultancy

Problem solving support

Implementing solution

Compile specialized software

Oversee purchases

DATA SHARING

Safe sharing, both within and outside of Iceland

Platform for cooperiation (NextCloud)

DATA STORAGE

Safe and efficient storage of reserach data

Certified storage of personal and sensitive data��ISO 27001 certified data management

COMPUTING POWER

HPC (CPU and GPU)

Directly connected to storage which can host large data

State of the art hardware for efficient analysis of data

Computing Power

Data storage

Consultancy

Data Sharing

6 of 20

Icelandic Jupyter Architecture Co-design by Use cases.

7 of 20

Interaction room

CoE RAISE Final Review – WP2 – Morris Riedel

focus input

focus input

smaller inputs

smaller inputs

[1] M. Book et al., Facilitating Collaboration in Machine Learning and High-Performance Computing Projects with an Interaction Room, IEEE E-SCIENCE 2022, https://doi.org/10.1109/eScience55777.2022.00093

Methodology that Dr. Matthias Book is working on

Revolves around splitting the system into four tables in an interaction room

Simplifies the importance of software systems

8 of 20

Interaction Room Remote Sensing

8/20

20XX

In-progress

Foundation Models for Generalist Geospatial Artificial Intelligence

Framework for building basic geographic models to accelerate the development and dissemination of climate and sustainability

Dr. Rocco Sedona

IAS, Jülich Supercomputing Centre (JSC)

Deputy Head of SDL AI and Machine Learning for Remote sensing

Trained for the use of data from satellite imagery.

2024

9 of 20

Interaction Room Lava Loba

9/20

20XX

Dr. Gro Birkefeldt Møller Pedersen

Running MrLavaloba for generating lavaflow models,

Volcanology expert at the Iceland Met Office

Worked as a PostDoc at the University of Iceland

In-progress

2024

10 of 20

Co-Design System Architecture

11 of 20

Motivation

  • Accessing to HPC cluster can be difficult

  • HPC systems can be complex
    • How do you submit scripts?
    • Infinyband no infinyband?
    • How do I use it in-general?

11/20

2024

12 of 20

Jupyter - JSC

12/20

2024

  • JupyterHub Web service running on HDF-Cloud at the Julich Supercomputing Centre.
    • Kubernetes cluster

  • Users use the Web service to run Jupyter Notebooks on HPC compute nodes.

  • Notebooks are deployed via a spawner that has its own environment on the compute nodes

[3] Jupyter. Jupyter Logo. Forschungszentrum Jülich, jupyter.jsc.fz-juelich.de. Accessed 3 Dec. 2024.

13 of 20

Jupyter - JSC Blueprint

20XX

14 of 20

Background UNICORE

14/20

2024

Separates into three different layers

  • Gateway: will be located on the UNICORE server

  • Unicore/X: will be located on the UNICORE server

  • TSI Layer: will be located on the node slogin2
  • Has been maintained for 25 years

  • Open-Source middleware for clients to connect to services in a secure manner (RESTful API)

[4] UNICORE. UNICORE Logo. unicore.eu

15 of 20

Background Unity IdM

15/20

2024

Offers a user-friendly way on setting up on managing user identities also compatible with Unicore

Offers many authentication settings, Oauth and SAML

Can be used to setup SSO connection

  • Makes it easier for users to access the resources in a secure manner

[5] Unity IDM. Unity IDM Logo. unity-idm.eu, www.unity-idm.eu.

16 of 20

UNICORE/X

Elja Specific Configuration

UTS Server

Landscape

UNICORE �Gateway Server

Elja Specific Configuration

UNITY IDM

UNICORE�Configuration

PORT XYZ

PORT XYZ

JUpy

Jypyter – JSC

Website

Internet

Elja LDAP�Server (Linux)�(no common user database)

UTS Server

2 landscape

LDAP information?

Add users, remove users for what is LDAP used

HI / UTS Configuration

LDAP

HI / UTS Configuration

firewall

UTS sever landscape

PORT XYZ

Software infrastructure Elja

Elja RMS

Slurm

Login node

UNICORE Target �System Interface (TSI)�(run as uid/setpriv)

Elja Specific Configuration�(usually runs as root)

Python�LavaLoba

Pytorch & TensorFlow�

PORT XYZ

Rocky Linux

Hardware infrastructure

EU �HPC �Systems

Elja

CPU

GPU

NetApp

Storage

Discs

Kubernetes Cluster

17 of 20

Plan of Implementation

18 of 20

Plan of Implementation

  1. Setting up UNICORTE/X and TSI on my personal computer

2. Setup connection to UNICORE/X from personal computer with TSI layer on Login Node

3. Set up connection from Unicore/X server to the TSI

4. Set up Unity IdM and connect to UNICORE/X

5. Set up connecting with Jupyter-JSC

6. Validate setup with Co-design use cases

18/20

2024

= To-do

= In-Progress

= Done

19 of 20

References

[1] M. Book et al., Facilitating Collaboration in Machine Learning and High-Performance Computing  Projects with an Interaction Room, IEEE E-SCIENCE 2022, https://doi.org/10.1109/eScience55777.2022.00093

[2] Jens Henrik Göbbert., Interactive HPC with Jupyterlab, EDIH-Workshop, part 3 Part3_JupyterLab_for_HPC/jupyter4hpc_part3_under-the-hood.pdf · main · jupyter4jsc / workshop-2023.10-jupyter4hpc · GitLab

[3] Jupyter. Jupyter Logo. Forschungszentrum Jülich, jupyter.jsc.fz-juelich.de. Accessed 3 Dec. 2024.

[4] UNICORE. UNICORE Logo. unicore.eu, https://www.unicore.eu/ . Accessed 3.Dec. 2024

[5] Unity IDM. Unity IDM Logo. unity-idm.eu, www.unity-idm.eu. Accessed 3 Dec. 2024.

19/20

2024

20 of 20

Takk fyrir mig

Sigurður Detlef Jónsson