CC-NIE Integration: ANSE (Advanced Network Services for Experiments)

California Institute Of Technology, Pasadena CA

Investigators

Abstract

Recent years have seen a dramatic evolution in the area of research and education networking, enabling faster data rates and novel services to user communities. The ANSE project integrates these new services, in particular dynamic bandwidth allocation (DYNES) and real-time network performance monitoring (perfSONAR), with the higher level workload management software deployed for the Large Hadron Collider (LHC), like PanDA and PhEDEx in the ATLAS and CMS experiments, respectively. A software defined networking approach, and in particular the use of dynamic bandwidth allocation capability, already available in ESnet and Internet2 networks, makes the network a truly manageable resource, on par with computing (CPU) and storage (disk and tape) resources. This approach enables smart optimization of resource utilization through co-scheduling and deadline scheduling algorithms within data and workload management systems. The ANSE system reacts to real-time conditions such as network events or node failure, as well as usage specific parameters like changes in data distribution priorities. Importance is given to the multi-domain aspects of networking, as required for a globally distributed system. While the primary target for ANSE is the LHC experiments, the ANSE project is clearly applicable to other data intensive science fields as well. Through the use of recognized standards such as the OGF NSI, ANSE is applicable beyond the use case of the LHC, and in particular to all distributed computing and data intensive applications of global scale.