A range-based SLA and edge driven virtual core provisioning in DiffServ-VPNs

We previously proposed a range-based service level agreement (SLA) approach and edge provisioning in DiffServ capable virtual private networks (VPNs) to customers that are unable or unwilling to predict the load between VPN endpoints exactly. With range-based SLAs customers specify their requirements as a range of quantitative values rather than a single one. Various suitable policies and algorithms dynamically provision and allocate resources at the edges for VPN connections. However, we also need to provision the interior nodes of a transit network to meet the assurances offered at the boundaries of the network. Although a deterministic guaranteed service (single quantitative value approach) provides the highest level of QoS guarantees, it leaves a significant portion of network resources on the average unused. We show that with range-based SLAs providers have the flexibility to allocate bandwidth that falls between a lower and upper bound of the range only, and therefore, take advantage of this to make multiplexing gain in the core that is usually not possible with a deterministic approach. But dynamic and frequent configurations of an interior device is not desired as this will lead to scalability problems and also defeats the purpose of the DiffServ architecture which suggests to drive all the complexities towards edges. We, therefore, propose virtual core provisioning that only requires a capacity inventory of interior devices to be updated based on VPN connection acceptance, termination or modification at the edges.