Architecture of our OpenStack deployment
by HEIG-Cloud
Tags:Posted on Wed, Dec 16, 2015
Architecture of our OpenStack deployment
This post is about the architecture of our OpenStack deployment at HEIG-VD. At maximum size our private cloud will run on 13 servers.
We chose a deployment that follows closely the standard installation guide. In total there are 13 servers with the following roles:
- One controller node
- One networking node
- Nine compute nodes
- Two storage nodes
When a server has a certain role it hosts a certain combination of OpenStack components. Roughly speaking, the controller node hosts central components like the message broker that is used by the OpenStack components to communicate with each other, the database, the authentication and authorization components and the user interface.
The networking node is responsible for attaching externally visible floating IP addresses to the virtual machines and provides virtual subnetworks and virtual routers.
A compute node contains a hypervisor that enables the creation of many virtual machines on that server.
A storage node provides its disk space in the form of virtual disks that can be attached to the virtual machines running in the compute nodes. To the guest OS in the VM they appear as any other disk and typically it will place a file system on them.
We have also created three separate networks:
- Management network (red): accessible only from the intranet
- Instance tunnels network (green): connects the compute nodes with the networking node
- External network (blue): makes the virtual machines accessible from the Internet (through the networking node) as well as the user interface (running on the controller node).