.png){kind=icon}
.png){kind=icon}
Return
Assembling multipath service function chains in substrate graphs using sharing instances and deep learning
.png){kind=icon}
PRE
AI
DOI：10.1016/j.engappai.2025.110900.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The assembly of Service Function Chains (SFCs) within substrate graphs has emerged as a crucial challenge in the context of network function virtualization. SFCs are ordered sequences of Virtual Network Functions (VNFs) that must be executed to meet specific service requirements. Efficient management of network resources in this domain can be achieved through the reuse of sharable VNF instances, which reduces resource consumption and operational overhead. As network environments evolve with increasing demands and dynamic complexities, multipath SFCs have gained attention as a promising approach to enhance fault tolerance and optimize resource utilization. Multipath SFCs allow traffic to traverse multiple distinct paths in the network, ensuring load balancing, redundancy, and improved performance. With this motivation, this study presents a multipath SFC assembly framework via VNF instance sharing in substrate graphs, implemented using deep learning algorithms. The proposed algorithm leverages the Soft Actor-Critic algorithm to embed multipath SFCs effectively. By dynamically mapping SFCs to paths with minimal bandwidth consumption and latency, our algorithm optimizes network resource usage. The resource allocation mechanism in proposed algorithm is designed to adapt flexibly to diverse request demands. Moreover, by utilizing reusable VNF instances, our algorithm further enhances resource efficiency. Also, the framework ensures that all nodes and links in the physical network satisfy resource constraints during the mapping process. Simulation results demonstrate that proposed algorithm significantly improves performance metrics under multipath SFC assembly with instance sharing in substrate graphs. Specifically, proposed algorithm outperforms state-of-the-art methods by improving bandwidth consumption by 1.04 %, reducing latency by 0.82ms, and increasing the SFC assembly rate by 2.75 %, demonstrating efficient network resource management.
Keywords:
Network function virtualization
Service function chains
Sharing instances
Substrate graphs
Deep learning
Journal
IF:
8
Papers:
5.4K
Citations:
3.5W