Cloud-Native Placement Strategies of Service Function Chains with Dependencies

Cloud services are now well established. Thanks to specific providers’ pioneering work, they offer on-site the benefit of predictability, continuity, and quality of service provided by virtualization technologies. In this context, SDN (Software Defined Networking) aims at providing tenant management of the transmission and various abstractions of the network infrastructure underlying the applications. Cloud platforms can also support virtualized network functions to complement the execution of online (web servers) or batch (compute or data-intensive) tasks. Scheduling and placing network functions into the cloud is a daunting task. One reason is that it requires time-consuming provisioning and configuration steps. This paper presents a generic framework that schedules network service function chains considering their internal dependencies. Toward this goal, our solution considers network functions’ placement, not their configuration. We are confronted with the general problem of defining the ordered sequence of service functions to be performed in a way that retains some criteria. Our framework considers dependencies within a service function chain but not between chains. We also perform experiments to highlight the benefits and properties of modeling work. The proposed generic framework can be instantiated with multiple multi-criteria decision supports and other techniques for placing final network functions. We conduct intensive experiments to find the best combination of strategies until the computing system exceeds 850 cores. Lessons learned are finally presented at the end of the paper.