With the increasing usage of cloud computing and dependence on a diverse set of distributed applications, users are reliant on consistent outcomes from a shared infrastructure. This drives the need for improved QoS guarantees for heterogeneous communication requirements over shared networks. The Recursive Inter-Network Architecture (RINA) is a fundamental programmable network architecture that provides a consistent model for supporting QoS across multiple layers. In this work we evaluate the performance outcomes provided by such polyservice RINA networks in conjunction with per-layer ΔQ-based resource allocation policies. ΔQ provides a resource allocation model able to enforce strict statistical limits on the maximum experienced losses and delays through the smart utilization of traffic policing and shaping strategies, together with an analytical pre-dimensioning of buffer thresholds. Our target scenario is a backbone network that prioritizes communications among geographically distributed datacentres using resources shared with best-effort background traffic. Results obtained with the RINASim simulation software show that a ΔQ-enabled RINA network can yield the desired absolute QoS guarantees to the assured traffic classes without negatively impacting the rest, unlike current MPLS-based VPN solutions.