The Platform-as-a-Service paradigm meets ATLAS: developing an automated analysis workflow on the newly established INFN CLOUD

The Worldwide LHC Computing Grid (WLCG) is a large-scale collaboration which gathers computing resources from more than 170 computing centers worldwide. To fulfill the requirements of new applications and to improve the long-term sustainability of the grid middleware, newly available solutions are being investigated. Like open-source and commercial players, the HEP community has also recognized the benefits of integrating cloud technologies into the legacy, grid-based workflows. Since March 2021, INFN has entered the field of cloud computing establishing the INFN CLOUD infrastructure. This platform supports scientific computing, software development and training, and serves as an extension of local resources. Among available services, virtual machines, Docker-based deployments, HTCondor (deployed on Kubernetes) or general-purpose Kubernetes clusters can be deployed. An ongoing R&D activity within the ATLAS experiment has the long-term objective to define an operation model which is efficient, versatile and scalable in terms of costs and computing power. As a part of this larger effort, this study investigates the feasibility of an automated, cloud-based data analysis workflow for the ATLAS experiment using INFN CLOUD resources. The scope of this research has been defined in a new INFN R&D project: the INfn Cloud based Atlas aNalysis facility, or INCANT. The long-term objective of INCANT is to provide a cloud-based system to support data preparation and data analysis. As a first project milestone, a proof-of-concept has been developed. A Kubernetes cluster equipped with 7 nodes (total 28 vCPU, 56 GB of RAM and 700 GB of non-shared block storage) hosts an HTCondor cluster, federated with INFN's IAM authentication platform, running in specialized Kubernetes pods. HTCondor worker nodes have direct access to CVMFS and EOS (via XRootD) for provisioning software and data, respectively. They are also connected to a NFS shared drive which can optionally be backed by an S3-compatible 2 TB storage. Jobs are submitted to the HTCondor cluster from a satellite, Dockerized submit node which is also federated with INFN's IAM and connected to the same data and software resources. This proof-of-concept is being tested with actual analysis workflows.