Security-Aware Cost Optimized Dynamic Service Function Chain Scheduling

Abstract

The idea of network function virtualization has emerged recently as a means of accelerating the deployment of middleboxes and network operations while also lowering deployment costs. Service function chaining provides network connectivity and steers traffic between the deployed VNF instances to provide a network service. However, integrating NFV within SFC scheduling introduces complexities, particularly in efficiently allocating resources to VNFs amidst dynamic network traffic and service demands. Optimization of VNF placement and scheduling is essential to minimize execution costs while meeting stringent Service Level Agreements (SLAs) and ensuring quality-guaranteed services. Moreover, the convergence of NFV and SFC scheduling brings for security challenges, including unauthorized access, data interception, and service disruption. Balancing optimization objectives with stringent security requirements poses a non-trivial task, emphasizing the need for prioritizing security in resource allocation and scheduling decisions. Meeting SFC deadlines is challenging due to dynamic network conditions, service demands, and resource allocation complexity. Failure to meet deadlines can lead to service quality degradation, SLA violations, and financial penalties. To address these challenges, the security and cost-aware SFC scheduling problem is formulated as an optimization problem. Moreover, a three-level security model is designed for both the VNFs and the physical machines in the NFV-enabled networks. Since the problem is NP-hard, we propose two heuristics named Particle Swarm Optimization-Based SFC Scheduling approach and Group Learning Particle Swarm Optimization-Based SFC Scheduling model that focus on optimizing the execution cost of SFC while meeting security and deadline requirements. The proposed models are compared with the existing SFC scheduling models. The effectiveness of the proposed scheduling approaches is evaluated through extensive simulations and it is shown that proposed scheduling approaches outperform the existing models in terms of average execution cost, security violation ratio, deadline violation ratio, service level agreement violation ratio, and average delay.