A rule-based modeling approach for network application availability assessment under dynamic network restoration scheme

The efficient and accurate assessment method for application availability is essential for network operators to make the resilience-cost trade-off policy. However, the backup path of application is dynamically discovered upon both other application requests and global network resource distribution under network restoration scheme, makes the current analytical network availability models difficult for assessing application availability effectively in large-scale network scenarios. This paper proposed a rule-based modeling approach to capture application resilience behavior and measure the uncertainties during dynamic restoration process. Firstly, we propose an evolution object formalism for modeling the attributes of both logical application and network infrastructure. Secondly, the network evolution condition model is proposed to describe the failure modes of individual components. Finally, the network evolution rule is proposed, which utilizes general graph algorithm and resource match function as local rules to determine application restoration resource contention outcomes during the state transition. Through simulations, the proposed approach evaluates application availability in terms of recovery time during system state transition controlled by the rules. The effectiveness of the proposed approach is verified to an actual optical network case under two typical network restoration schemes and compare its accuracy against two analytical approaches based on RBD (Reliability Block Diagram) and Markov Chain. The sensitivity results reveal that link capacity has a more significant impact on application availability than link failures.