Congestion-based API for cloud and WAN resource optimization

Cloud services span across geographically distributed data centers (DCs), between which a large amount of traffic is transfered. These DCs are interconnected via a wide area network (WAN) through statically set up Virtual Private Networks (VPNs), so they do not provide the flexibility offered by the modern cloud services. Moreover, the traffic over the WAN has many constraints (e.g., bandwidth, delay, loss), which brings concerns about the quality of the interconnections between DCs. By decoupling the forwarding and the control plane and centralizing the control plane in a network controller aware of the network status, the Software Defined Networking (SDN) paradigm provides the ability to program the network, establish VPNs on demand, and monitor the network traffic to maintain a good quality of service (QoS) for the cloud tenants. This paper presents an architecture based on an orchestrator of both cloud and WAN resources, which makes possible the dynamic establishment of VPNs between DCs, and the monitoring of the network traffic through a dedicated application programming interface (API) on the WAN controller. This allows to take into account the network status in the deployment stategies of computing elements like Virtual Machines (VMs) over the geographically distributed DCs. A Proof of Concept (PoC) has been developed to experimentaly demonstrate the proposed architecture and API.