This article talks about how Kubernetes has emerged from container orchestration platform to manage complex workloads in AI and Machine Learning Stacks, Managing containers in NFV architecture and handling hardware GPU resources.
Kubernetes works on the principle of assigning IP addresses to pods, called as “IP-per-pod” model. The IPAM (IP address management) task is left to third party solutions. Some of these solutions include Docker networking, Flannel, IPvlan, contive, OpenVswitch, GCE and others.
The Kubernetes architecture consists of master node, replication controller in addition (or conjunction) to nodes used to host the pods. Before we go ahead, here is a review of Kubernetes terms.