Demonstration: Network as Resource in Big PanDA

15/06/2015 13:30-14:00
Demonstration area
PanDA is the workload management system used by thousands of physics users in the ATLAS experiment at the LHC to analyze the multi Petabyte detector data. PanDA gives the appearance of a single system, but the reality is that underneath there is a worldwide mesh of heterogeneous data centers with different processing, storage and also network connectivities. Historically, the data centers were hierarchically grouped into clouds according to geographical boundaries following the layout of national research networks. However, the offerings of the research networks bandwidth today has progressed so far that the original boundaries need to be relaxed and we have the possibility to use the network as a resource in order to optimize our workflows, the same way we do for storage or CPU. This demo will give an insight into the PanDA workload management system and explain how it is dynamically using network metrics to optimize job brokerage and hence reduce user waiting time.