AUDITOR: Accounting data handling toolbox for opportunistic resources

Abstract

Increasing computing demands and concerns about energy efficiency in high-performance and high-throughput computing are driving forces in the search for more efficient ways to use available resources. Sharing resources of an underutilised cluster with a high workload cluster increases the efficiency of the underutilised cluster. The software COBalD/TARDIS can dynamically and transparently integrate and disintegrate such resources. However, sharing resources also requires accounting. AUDITOR (Accounting Data Handling Toolbox for Opportunistic Resources) is a modular accounting ecosystem that is able to cover a wide range of use cases and infrastructures. Accounting data are gathered via so-called collectors, which are designed to monitor batch systems, COBalD/TARDIS, cloud schedulers, or other sources of information. The data is stored in a database, and access to the data is handled by the core component of AUDITOR, which provides a REST API along both Rust and a Python client libraries. So-called plugins can take actions based on accounting records. Depending on the use case, one simply selects a suitable collector and plugin from a growing ecosystem of collectors and plugins. To facilitate the development of collectors and plugins for yet uncovered use cases by the community, libraries for interacting with AUDITOR are provided.