NFV Optimization Algorithm for Shortest Path and Service Function Assignment

Abstract

Our paper focuses on the concept of Network Function Virtualization (NFV): the implementation of requests consisting of various service functions on servers located in data centers. This paper attempts to minimize both the cost of routing and service function assignment of requests from source to destination node on a network. This problem falls under the class of Integer Linear Programming (ILP) which is NP Hard and cannot be solved in polynomial time. Towards developing a solution, we propose to split the problem into two separate optimization subproblems: shortest path routing and service function assignment. We utilize Dijkstra's Shortest Path algorithm and a Greedy method for service function assignment to propose a new heuristic algorithm that minimizes the total cost of routing and service functions assignment. We also analyze the run-time complexity of our proposed heuristic algorithm. Our experimental results suggest that our proposed algorithm matches the optimal ILP solution within acceptable limits.