STAR-Scheduler: A Batch Job Scheduler for Distributed I/O Intensive Applications

We present the implementation of a batch job scheduler designed for single-point management of distributed tasks on a multi-node compute farm. The scheduler uses the notion of a meta-job to launch large computing tasks simultaneously on many nodes from a single user command. Job scheduling on specific computing nodes is predicated on the availability of user specified data files co-located with the CPUs where the analysis is meant to take place. Large I/O intensive data analyses may thus be efficiently conducted on multiple CPUs without the limitations implied by finite LAN or WAN bandwidths. Although this Scheduler was developed specifically for the STAR Collaboration at Brookhaven National Laboratory, its design is sufficiently general, it can be adapted to virtually any other data analysis tasks carried out by large scientific collaborations.