Design of Scheduler Plugins for Reliable Function Allocation in Kubernetes
Rui Kang, Mengfei Zhu, Fujun He, Takehiro Sato, Eiji Oki · 2021
The reliability of virtual network can be increased by allocating virtual network functions (VNF) to suitable locations. The VNF placement problems are formulated as optimization models with different objectives. The models are solved by optimization software and heuristic algorithms. The allocation results obtained by the models are used to allocate the VNFs to nodes. Since different users have different objectives, it is necessary to allocate different groups of VNFs by using different models. Existing tools did not provide a method to connect multiple optimization models with Kubernetes. We implement function scheduler plugins cooperating with multiple reliable function allocation models in Kubernetes, which is a system for automating deployment, scaling, and management of containerized applications. Demonstration validates that the plugins allocate functions by using the allocation results obtained by the model automatically and run service functions correctly.