Analysis of Unavailability in Middleboxes With Double-Capacity Multiple Backup Servers Under Shared Protection
Naohide Wakuda, Ryuta Shiraki, Eiji Oki · IEEE Transactions on Network and Service Management · 2025
Middleboxes play a critical role in network operations, providing various network service functions, and can be implemented as software on general-purpose servers through network function virtualization technology. The unavailability of middlebox functions is an essential metric of the overall quality of network services. This paper proposes an analytical model that calculates the unavailability of middlebox functions, where multiple backup servers protect one or more functions, and one backup server can recover at most two functions simultaneously, which we call the double-capacity multiple-backup model. While the single-capacity multiple-backup model, where each backup server can recover at most one function, has been studied, the double-capacity multiple-backup model has not been addressed. The proposed double-capacity multiple-backup model allows for load balancing among backup servers. The proposed model can have two different workload strategies: load-persistent (LP) and load-distributing (LB). We use a Markov chain to analyze state transitions and develop equilibrium-state equations, providing a method to compute function unavailability for each strategy. Numerical results observe that these two strategies have almost the same unavailability. Since the LB strategy incurs additional operational costs compared to the LP strategy, the LP strategy is preferable. We also find that the double-capacity multiple-backup model reduces unavailability by 7.21–81.9% compared to the single-backup model in the examined cases.