Infrastructure of AI4PublicPolicy services pt.1

The AI4PublicPolicy services are built on top of the multi-layered EOSC Compute Platform

Its main layers are the Federated Resource Providers layer, the Compute and Data Federation layer, the Platforms layer, and the vertical Service Management Tools layer.

The vertical Service Management Tools layer includes Helpdesk, Monitoring, Accounting, Operations Management, Security and Incident response, and Software Quality Assurance services.

The Federated Resource Providers layer provides an Infrastructure-as-a-Service (IaaS) hybrid infrastructure from academic and commercial providers to run and/or host research applications and data. Its components are provided by cloud providers and High-Performance Computing (HPC) centres from different European countries.

The Compute and Data Federation layer provides advanced solutions for facilitating common and recurring tasks such as data management and transfer, virtual infrastructure orchestration and software distribution.

The Platforms layer provides a higher abstraction layer on top of the previous one, offering generic Platform-as-a-Service (PaaS) services such as interactive notebook environment, machine learning and deep learning applications for data analysis, scalable big data tools, and workload management tools.

The infrastructure provides technologies to allocate cloud and HPC resources like computing, data storing, and sharing to facilitate the management of containers for the execution of the AI models created in the project. These resources are offered through the European Cloud Initiative. EGI FedCloud services are used where applicable.

The infrastructure is populated by data, applications, scripts, AI models etc. from project pilots and application developers.

In this first part of the blog posts about the infrastructure of AI4PublicPolicy services, the Service Management Tools layer and the Federated Resource Providers layer are presented:

Service Management Tools

One of the main services provided by the Service Management Tools layer is access to datasets and applications hosted on the Compute Platform, which can be defined through Check-in.

Authentication and Authorization: Check-in is the authentication, authorisation and user management service for the EGI infrastructure. It enables users to access EGI, and third-party services (web and non-web based), using existing credentials managed by the Identity Providers (IdPs) of their home organisations. The Check-in service enables users to manage their accounts from a single interface, link multiple accounts/identities together and access services based on their roles and Virtual Organisation (VO)/group membership rights.

Federated Resource Providers

The Federated Resource Providers layer is the bottom layer of the infrastructure providing computing and data resources both on cloud or HPC.

Cloud providers: Infrastructure-as-a-Service (IaaS) Cloud is provided by research cloud providers from many European countries and is based on OpenStack. It can be accessed through Check-in via a web dashboard, API and Command Line Interface (CLI). They allow the provision of low-level storage such as block storage and object storage as well as Virtual Machines (VMs) and Kubernetes cluster among other things.

HPC: High Performance Computing (HPC) supports highly optimized applications that need massively parallel computing with low latency and high bandwidth interconnection networks such as complex computational problems using tightly coupled parallel processing simulations, analysis of large datasets or AI workloads. It provides highly optimized computing systems that deliver large amounts of parallel computing power to run such applications and offers a managed service where you can find a fully operational environment where to submit your jobs. In most HPC systems, user access is performed via the Secure Shell Protocol (SSH) to a set of login nodes where users can interact with the system and submit jobs for their execution. In those providers who deliver HPC hardware via IaaS interfaces, an HPC system can be deployed using tools such as Elastic Cloud Compute Cluster.

Stay tuned for part 2 of the Infrastructure of AI4PublicPolicy services blog series, where the Compute and Data Federation layer and the Platforms layer will be highlighted.