The computing cluster at CNAF is capable of supporting local, Grid and Cloud resource requests on both local and remote environments.
Currently half of the farm computing power comes from leased CINECA nodes (PRACE Tier-0, located in Bologna), connected to CNAF through 400Gbit connection. A smaller fraction of the resources come from another INFN center, the Bari ReCaS facility.
Jobs running on the farm have direct access to files: the disk storage system is organized through several file systems directly mounted on the Worker Nodes.
Tipically, each experiment has at least a dedicated queue, and the computing resources are centrally managed by a unique batch system. The allocation of resources is based on a Hierarchical Fairshare scheduling: it is directly proportional to the assigned share of the experiment the submitter of the job belongs to, and inversely proportional to the historical resource usage. This strategy allows to maximize CPU utilization, so that the Tier-1 farm runs at 100% of its power most of the time.
On average, 100000 batch jobs are executed every day, and computing resources are used 24×7 by all the scientific experiments.
The bulk of the computing resources (Worker Nodes, User Interfaces, grid services) and the user database are operated by the Farming group of Tier-1 data center unit.