Dynamic Service Chain Deployment Method with High Reliability Awareness

Aiming at 5G virtual network SFC more faults cause by virtualized resource slicing, and existing studies have low reliability evaluation accuracy, lack differentiated reliability guarantee, this paper put forward a high reliability-aware dynamic SFC deployment scheme; first, the scheme build reliability index based on device usage rate and surrounding reliability coefficient, set differentiated reliability needs according to VNF characteristics, and limit link length to cut backup resource cost, in VNF deployment phase, we propose a node ranking algorithm based on node location and reliability, use Google PageRank core idea to define node importance and guide deployment, link mapping adopt reliability-based method, it sort links with high reliability needs, use K-shortest path algorithm to filter, balance reliability and delay finally; to further ensure reliability, we design a differentiated backup scheme based on VNF and link importance, VNF with importance above average use dedicated backup, below average use shared backup, we introduce unit resource reliability improvement rate index to optimize link backup, it can make backup more efficient without too much resource waste; we build optimization model based on deep reinforcement learning, it interact continuously with virtual layer and underlying environment, maximize cumulative reward to get best deployment strategy, realize joint optimization of load balance and VNF reliability; simulation results show that, the proposed DQN-RDA algorithm in 50 physical nodes network, request acceptance rate keep around 90%, compare with existing algorithms, it perform better in fault rate, failure loss and backup resource consumption, this algorithm can well adapt to virtual network dynamic changes, it can meet high reliability needs of various applications.