Geography-based Service Function Path (SFP) Provisioning in Inter- Datacenter Elastic Optical Networks

摘要

Inside a service function chain (SFC). traffic flow follows a certain route, namely a service function path (SFP), to travel through each service function (SF) entity. A SFP consists of.several end-to-end segments, whose source and destination are named anchor node (AN). SFs are located in multiple datacenters (DCs), and inter-DC light-paths need to be provisioned between separated SFs. In this paper, we introduce geography information of optical nodes and DCs, define special geographic distance between ANs in inter-DC elastic optical networks (EONs). Then following minimal geographic distance principle, we propose a geography-based SFP provisioning solution, which contains two heuristic algorithms, named geography-based shortest path and first-fit algorithm (GSP-FF) and geography-based k-shortest paths and first-fit algorithm (GK-FF). These algorithms can compress AN selection procedure extremely in fixed time, which cost little time for the AN selection of resource allocation. And benchmark algorithm use Dijkstra shortest path calculation and first-fit FS selection to allocate IT resources in DCs and FS resources in EONs. Then GSP-FF and GK-FF are proposed to provision SFPs efficiently. In our simulation, we compare our proposed algorithms with benchmark algorithm deeply on blocking probability, running time, average hops, average geographic distance, el al. under different traffic load and other simulation environment. We also analyze the trend and reason for the performance difference among these algorithms. According detailed evaluation, simulation proves that the proposed algorithms in this paper could use geographic information efficiently, and achieve lower blocking probability with lower running time compared with the benchmark algorithm.