CERN Accelerating science

Batch Share Management Tool

Date published: 
Monday, 1 September, 2014
Document type: 
Summer student report
T. de Kock
Project goals: One of the key computing services at CERN is the central batch system. The central batch service currently consists of around 4000 machines with thousands of users, divided over more than 250 groups. The central batch service system runs IBM/Platform LSFr. Additionally, a pilot service based on HTCondor is currently being set up. In a batch system the worker nodes are shared among the users over time. The central batch service uses fair-share scheduling, which ensures that, over a historic time window, a group of users can use the capacity (share) that is assigned to them. Each share on the system is allocated to a group of users. Compute coordinators can delegate capacity, as well as the control of this capacity to subgroups. This situation is shown in figure 1.