This guide provides the steps needed to bring a Universal Compute Platform cluster online. Universal Compute Platform leverages Kubernetes and Docker to deploy and manage the delivery of applications to the customer premises.
The following figure depicts the three physical host boxes required for Universal Compute Platform, with ports mapped as follows:
As an option, the system leverages VRRP (Virtual Router Redundancy Protocol) to provide support for both high-availability and load balancing, supported by an NGINX engine. All service operations to the cluster should be directed to the corresponding VRRP IP so that the load balancing logic can direct the request to the best node.
Deployment configuration requirements vary over different applications deployed into the Universal Compute Platform. One main requirement in the establishment and operation of the cluster is the Inter-Cluster Connection. This connection operates as the backplane between nodes in the cluster. This backplane carries all the synchronization data between nodes for both component and data states. It is a best practice to deploy the interface as a segregated 10 Gbps inter-connect (separate switch port), allowing for the best performance in synchronization between nodes.