Monitoring and analyzing the execution of a workload is at the core of the operation of data centers. It allows operators to verify that the operational objectives are satised or detect and react to any unexpected and unwanted behavior. However, the scale and complexity of large workloads composed of millions of jobs executed each month on several thousands of cores, often limit the depth of such an analysis. This may lead to overlook some phenomena that, while not harmful at a global scale, can be detrimental to a specic class of users. In this paper, we illustrate such a situation by analyzing a large High Throughput Computing (HTC) workload trace coming from one of the largest academic computing centers in France. The Fair-Share algorithm at the core of the batch scheduler ensures that all user groups are fairly provided with an amount of computing resources commensurate to their expressed needs. However, a deeper analysis of the produced schedule, especially of the job waiting times, shows a certain degree of unfairness between user groups. We identify the conguration of the quotas and scheduling queues as the main root causes of this unfairness. We thus propose a drastic reconguration of the system that aims at being more suited to the characteristics of the workload and at better balancing the waiting time among user groups. We evaluate the impact of this reconguration through detailed simulations. The obtained results show that it still satises the main operational objectives while signicantly improving the quality of service experienced by formerly unfavored users.

A common characteristic to major physics experiments is an ever increasing need of computing resources to process experimental data and generate simulated data. The IN2P3 Computing Center provides its 2,500 users with about 35,000 cores and processes millions of jobs every month. This workload is composed of a vast majority of sequential jobs that corresponds to Monte-Carlo simulations and related analysis made on data produced on the Large Hadron Collider at CERN. To schedule such a workload under specic constraints, the CC-IN2P3 relied for 20 years on an in-house job and resource management system complemented by an operation team who can directly act on the decisions made by the job scheduler and modify them. This system has been replaced in 2011 but legacy rules of thumb remained. Combined to other rules motivated by production constraints, they may act against the job scheduler optimizations and force the operators to apply more corrective actions than they should. In this experience report from a production system, we describe the decisions made since the end of 2016 to either transfer some of the actions done by operators to the job scheduler or make these actions become unnecessary. The physical partitioning of resources in distinct pools has been replaced by a logical partionning that leverages scheduling queues. Then some historical constraints, such as quotas, have been relaxed. For instance, the number of concurrent jobs from a given user group allowed to access a specic resource, e.g., a storage subsystem, has been progressively increased. Finally, the computation of the fair-share by the job scheduler has been modied to be less detrimental to small groups whose jobs have a low priority. The preliminary but promising results coming from these modications constitute the beginning of a long-term activity to change the operation procedures applied to the computing infrastructure of the IN2P3 Computing Center.