Performance guarantee aware orchestration for service function chains with elastic demands

摘要

In this paper, we study how to provide optimal performance guarantee in multi-resource scenarios, such as Network Function Virtualization (NFV), where Virtual Network Function (VNF) resource demands of one Service Function Chain (SFC) are elastic and correlated with the end-to-end throughput of SFC. Previous works such as Dominant Resource Fairness (DRF) achieve this goal only if the demands are already orchestrated to dedicated resource instances. However, how to orchestrate SFC to VNF instances is a key issue in NFV. Few previous works have investigated the relationship between them. In this paper, we consider the problem of SFC orchestration subject to DRF allocation scheme so as to find the orchestration scheme with optimal performance guarantee. We analyze it and model it as SFC Performance Guarantee (SFC-PG) problem, followed by a Mixed Integer Linear Programming (MILP) formulation. We design a heuristic Performance Guarantee Aware Orchestration (PGAO) to solve it in large scale cases. We conduct simulations with a variety of circumstances and parameters, and the results demonstrate that PGAO achieves better performance compared to existing solutions.