Dynamic Flow Migration: A Model-Based Optimization Approach

In this chapter, a dynamic flow migration problem for embedded services is studied, to guarantee per-service average E2E delay in the presence of time-varying traffic. To address the trade-off between load balancing and migration cost in flow migration, a multi-objective mixed integer optimization problem is formulated. An optimal MIQCP solution with MIQCP problem transformation and optimum mapping provides an optimal solution to the optimization problem. Due to the NP-hardness of the reformulated MIQCP problem, a low-complexity heuristic solution is used to find a time efficient suboptimal solution. Numerical results are presented to demonstrate the aforementioned trade-off between load balancing and migration cost in different settings, compare the E2E delay performance with and without flow migration, and evaluate the effectiveness and efficiency of the heuristic solution.