Management of computational infrastructure is a complicated task which, often employs user workloads delivery across multiple clusters. Criteria for such tasks distribution may vary: priority, transport costs, utilization of data, node capabilities, etc.

Such process happens to tasks devoted to the simulation and analysis of the results of high-energy physics experiments at CERN. For task distribution on massive data streams obtained during ATLAS experiment, “Production ANd Distributed Analysis system” (PanDA) was developed. It performs management of workloads delivery and execution in a geographically distributed cluster environment. This paper is devoted to the deployment of PanDA server in a private cluster setting.

This paper presents architecture and its implementation that allows, to run and embed PanDA system into existing computational solutions. It consists of a container, that isolates PanDA server its dependencies and environment from other system processes and an embedded Web interface which simplifies task management for end-users. In other words, our approach is focused on PanDA system deployment speed up by means of security layer simplification, containerization and stateless client web service implementation. System was tested on a heterogeneous geographically distributed Azure cloud nodes.