A Reliability-Aware Network Service Chain Provisioning With Delay Guarantees in NFV-Enabled Enterprise Datacenter Networks
Long Qu,Chadi Assi,Khaled Shaban,Maurice J. Khabbaz
DOI: https://doi.org/10.1109/tnsm.2017.2723090
2017-09-01
IEEE Transactions on Network and Service Management
Abstract:Traditionally, service-specific network functions (NFs) (e.g., Firewall, intrusion detection system, etc.) are executed by installation-and maintenance-costly hardware middleboxes that are deployed within a datacenter network following a strictly ordered chain. NF virtualization (NFV) virtualizes these NFs and transforms them into instances of plain software referred to as virtual NFs (VNFs) and executed by virtual machines, which, in turn, are hosted over one or multiple industry-standard physical machines. The failure (e.g., hardware or software) of any one of a service chain’s VNFs leads to breaking down the entire chain and causing significant data losses, delays, and resource wastage. This paper establishes a reliability-aware and delay-constrained (READ) routing optimization framework for NFV-enabled datacenter networks. READ encloses the formulation of a complex mixed integer linear program (MILP) whose resolution yields an optimal network service VNF placement and traffic routing policy that jointly maximizes the achieved respective reliabilities of supported network services and minimizes these services’ respective end-to-end delays. A heuristic algorithm dubbed Greedy- ${k}$ -shortest paths (GSP) is proposed for the purpose of overcoming the MILP’s complexity and develop an efficient routing scheme whose results are comparable to those of READ’s optimal counterparts. Thorough numerical analyses are conducted to evaluate the network’s performance under GSP, and hence, gauge its merit; particularly, when compared to existing schemes, GSP exhibits an improvement of 18.5% in terms of the average end-to-end delay as well as 7.4% to 14.8% in terms of reliability.
computer science, information systems