Resilient Multi-Tenant Network Programmability for Adaptive Service Placement

This paper presents a resilient multi-tenant network programmability framework for adaptive service placement. In the increasingly dynamic world of network technologies, the ability to perform code updates quickly and reliably for multiple tenants is crucial to address the demands of modern network management and security in programmable networks. This paper presents an approach for resilient multi-tenant code updates focusing on adaptive network service placement. By extending our previously presented Orchestrator for Multi-Tenant Programmable Data Plane (PDP) Code Updates (OMuProCU), we not only enable seamless data plane code updates, but also ensure holistic and consistent network state updates without disrupting operations.Our approach adds state synchronization across multiple orchestrators as well as their components and addresses the coordination of network behavior in response to PDP code updates. We describe scenarios for accelerated containerized network functions migration in edge computing and 5G networks, which significantly benefit from our enhanced orchestration, and evaluate the benefits of resilient network state management in the improved orchestrator. Furthermore, the Quality of Service (QoS) of use cases in 5G networks and how they can benefit from our orchestrator are described. Our case studies, focusing on edge computing networks and integration into 5G infrastructures, outline OMuProCU's pivotal idea in fulfilling stringent QoS requirements emanating from varied use case demands, ranging from ultra-low latency to high throughput and reliability. This is evaluated by a Proof-of-Concept traffic application and a comparison with deploying CNF in a private and public cloud besides the accelerated CNF deployed by our framework. The solution aims to increase network resilience and flexibility, paving the way for adaptive, reliable state changes in an era where network programmability and multi-tenancy are in the forefront. This is a preprint version of the manuscript.