Grids generally rely on a complex interconnection of IP domains that may exhibit changing performance characteristics and may offer different quality of service facilities. However the end to end performances the heterogeneous mix of grid flows get from the network may affect the overall grid infrastructure performances, utilization level and impact each individual application's performance. In this paper we examine the case of the eToile french grid testbed based on a research network infrastructure that proposes a set of Differentiated Services with various end to end performances. We present the QoSinus service that aim to dynamically allocate the network resources to Grid flows in order to match their specific requirements under different load conditions. The QoSinus service is based on the programmable network approach. We show how a biomedical application distributed over the grid may benefit from such an advanced and controlled network service.