Welcome address from the IT department head

• Simone Campana (CERN)

Opening Session

• Alberto Pace (CERN)

Opening Lecture: The need for IT Service in accelerator and particle physics

• Sebastien Ponce (CERN)

Starting from accelerator and particle physics, we'll try to see what are the needs of the experiments and accelerator people in terms of IT services.

Database Services (part 1 of 4) - Introduction to DBoD - Short intro about databases - Presentation of the DBoD service - How to create a database in DBoD - How to connect to my DBoD

Storage (part 1 of 2) and pre-requisites This two-part lecture series provides an overview of the various storage services at CERN. We will look into the motivation behind our large scale storage systems, cover some fundamentals and the design principles of the storage systems we've developed and use. We will also look into some practical use cases covering the ecosystem of the many storage systems we run in the IT department. This should serve as a basis for choosing the correct storage services for the applications you would develop and practical considerations into utilizing storage effectively.

Storage (part 2 of 2)

Project Management and documentation In this session we will explore the solutions for project management and effective software development. We'll start by demonstrating how to plan and track project progress using Jira, available at https://its.cern.ch. Next, we'll highlight how GitLab Pages can be utilized to store and share information, whether for internal use or a broader audience, showcasing instances of technical documentation delivered to end-users. Finally, we'll explore Confluence for documentation storage, providing a walkthrough of its features and real-world examples of its use.

Student self presentation

Modern Application Development & Deployment “I need to develop an application X for user community Y, which will need to be run and maintained over time” 2. Application development a. This part of the session will explore the multiple types of applications, how to leverage version control system Gitlab and it's CI to have modern application deployment. This is organized as a workshop and will include a hands-on experience covering the best practices to develop containerized applications and strategies for deploying them. ------ Participants will begin by exploring various application types and learning how to leverage GitLab's version control and CI pipelines for efficient deployment. Through hands-on exercises, attendees will develop a simple application, while learning key concepts such as: - Best practices for developing containerized applications. - Writing DockerFiles for application deployment. - Setting up continuous integration (CI) workflows to automate testing and build and publish Docker images.

Creating and managing websites at CERN An introduction to website provisioning and hosting at CERN with the services provided by IT. We will focus on creating websites with WordPress and go into the details of the WordPress service at CERN.

Core compute services (part 1 of 2) An in depth set of use case where IT services are heavily used for physics, analysis and engineering applications. In part I of the series we will show use cases for Openstack, Linux and virtual-machine based configuration management.

Core compute services (part 2 of 2) An in depth set of use case where IT services are heavily used for physics, analysis and engineering applications. In part 2 of the series we will explore HTCondor, the high-throughput compute platform used for batch computing.

Configuration & Infrastracture as Code

• Daniel Juarez (CERN)

IT central configuration management services for virtual and physical hosts.

Database Services (part 2 of 4) - DBoD maintenance exercises - What are the typical tasks to be performed as DBoD owner Exercises on: - Cloning mechanism - Upgrades - TLS certificates

Data Analysis Techniques using SWAN and REANA (part 1 of 3) In this first session, we will give an overview of the SWAN service. This will include the following points: - Interface: classic and JupyterLab - Creation of projects, notebooks and terminals - Integration with CVMFS for software provisioning - Integration with EOS for storage and CERNBox for sharing - Use of GPUs - Connection to Spark clusters Moreover, we will give a live demo that participants will be able to follow along and get familiar with the basic features of SWAN.

Authentication and authorization (Part of "Software development and hosting" track) In this class, we will see how to: - Enable authentication with CERN SSO - Define role-based authorization for our applications - Get tokens, and use them access APIs

Authentication and authorization (Excercises) (Part of "Software development and hosting" track) In this class, we will see how to: - Enable authentication with CERN SSO - Define role-based authorization for our applications - Get tokens, and use them access APIs

Services for Machine Learning applications (part 1 of 3) This session will introduce the different phases in a ML lifecycle, and how IT services can help in each of the parts. In particular, it will: * Overview of ML and use cases, containerization and how it helps out in defining single units of computation, isolate custom software environments, and ensure sustainability for reproducible results * Demo how cloud native environments (Kubernetes and its ecosystem) can help manage those units of computation and scale them out to large amounts of resources * Provide an example of how to scale out using both on-premises and public cloud resources, and when this might be useful and cost effective

Lightning talks - Evangelos Katralis - GPU acceleration of a Particle in Cell (PIC) code - Jorge Belmonte - File Conversions and Batch Jobs Through Cluster Computing for the PLM Platform - Andrzej Nowicki - AI Meets Databases: How to build an image search using pgvector extension of PostgreSQL

Services for Machine Learning applications (part 2 of 3) This session will focus on the infrastructure and low level tools required to efficiently deploy machine learning applications. In particular, it will cover: * The different data types and how they can impact ML workloads, as well as support in different types of hardware and software libraries * Key differences between CPUs and GPUs and how they impact ML workloads (training and serving) * The available techniques in IT services for GPU sharing and partitioning. In particular, it will cover how applications can build on the existing Kubernetes service to simplify these operations * Hands-on exercises on using GPUs for different types of work

Database Services (part 3 of 4) - Oracle Database - Introduction of the Oracle Database service - Resource portal as a way to manage Oracle database users - needed e-groups - Other tooling provided by the team - Session Manager - How to connect to the Oracle database - Where to get the client? - What is the tnsnames.ora file?

Services for Machine Learning applications (part 3 of 3) This session will focus on available ML techniques for distributed training of models, hyperparameter optimization and model service. In particular, starting from a well known use case it will demonstrate: * How to go from a script, to a docker image training on a single node, to a distributed training setup with multiple nodes * How to do hyperparameter optimization, which kind of optimizers are available, how to monitor the workloads and how to publish the models * How to serve models in production, at scale, with a simple http entrypoint or embedding the model in an application

Introduction to registry.cern.ch and Kubernetes

Deploying applications (part 1 of 2) In this lecture, we will understand the difference between IaaS, PaaS and SaaS. Then, we will learn how to deploy custom and off-shelf applications to OKD PaaS.

Deploying applications (part 2 of 2) Exercises regarding how to deploy custom and off-shelf applications to OKD PaaS.

Data Analysis Techniques using SWAN and REANA (part 2 of 3) In the second session of this series, we shall present REANA reusable and reproducible analysis platform. REANA allows researchers to structure their data analyses by means of declarative workflow languages (CWL, Snakemake, Yadage) and run containerised data analysis pipelines on remote compute clouds (Kubernetes, HTCondor, Slurm). In the first part of this session, we shall discuss the notions of computational reproducibility and reusability, underlying the importance of encapsulating the original computing environments by means of containers and documenting the steps necessary to arrive at results. We shall provide a brief introduction to declarative workflow languages and discuss its pros and cons when compared to imperative analysis code programming. In the second part of this session, the participants will familiarise themselves with the REANA platform by means of running a simple analysis example. We shall use the [https://reana.cern.ch](https://reana.cern.ch/) instance at CERN to run a [RooFit demo example](https://github.com/reanahub/reana-demo-root6-roofit/).

Data Analysis Techniques using SWAN and REANA (part 3 of 3) In the third session of this series, we will propose short exercises using SWAN and REANA to cover more data analysis examples and use cases. The session will be split in two parts, one for each tool, where participants will be able to work on the exercises and get assistance from the lecturers.

Database Services (part 4 of 4) - DBoD maintenance exercises - What are the typical tasks to be performed as DBoD owner Exercises on: - Cloning mechanism - Upgrades - TLS certificates

Application security Short introduction to best practices for secure development, testing and deployment - Three golden rules for system security - Software security, typical vulnerability types - How security analysis tools can help - Introduction to penetration testing - Deployment security best practices

Closing Session

• Alberto Pace (CERN)