Adaptive management of connections to meet availability guarantees in SLAs

Today's backbone communication networks serve a wide range of services with different availability requirements. Each customer has a contract, denoted a Service Level Agreement (SLA) which specifies the availability requirement over the contract period. In the literature, different provisioning strategies to establish connection arrangements capable of meeting a statistical asymptotic availability for the different customers have been proposed. In reality, the SLAs specify guarantees on the interval availability which may deviate significantly from the asymptotic availability. This paper proposes to use an adaptive strategy to manage which connections are affected by failures and maximize the compliance with the SLAs. Different policies for management of connections from the same class with equal requirements and connections with different requirements are proposed. These policies are evaluated and compared with the traditional provisioning policies in a simulation study. The results show that adaptive management can significantly reduce the risk of violating the SLAs in several scenarios.