Powered by OpenStack [Version 11.07.11]
OpenStack provides a number of network models to choose from when designing a deployment; all are described below. However, for this reference architecture, the High Availability DHCP model is described in more detail than the full list.
Flat Network Model – A network administrator specifies a subnet from which all the virtual machines pulls IP addresses from a pool of available fixed addresses.
Flat DHCP Network Model – The server that runs nova-network is a gateway to the compute nodes running virtual machines.
VLAN Model – The server running virtual machines (a compute node) creates a VLAN and a bridge for each project or tenant, and users access their VMs through a special VPN that must be created.
High-Availability FlatDHCP Model (the Rackspace Cloud Builders Default) - Each compute host does Network Address Translation (NAT), DHCP, and acts as a gateway for all of its own virtual machines.
Rackspace Cloud Builders deploy a High Availability FlatDHCP networking model provided by OpenStack. This network model requires that the nova-network software is installed and configured on each server that is running nova-compute. The purpose of spreading the network service across multiple servers is to localize the failure domain to each nova-compute node. In a scenario where a nova-compute server is taken offline for any reason including maintenance, only the virtual instances on that server will be affected. All other instances in the private cloud will continue to serve traffic through their own network service. This is depicted in the example below.
Figure 3 High-Availability DHCP Networking Model
This section describes the types of networks you need to configure to work with an OpenStack deployment. These contain best practices for both conserving network resources and ensuring that network administrators understand the needs for networks and public IP addresses for accessing the APIs and VMs as necessary. It offers recommendations and required minimum sizes.
Management Network (RFC1918 IP Range, not publicly routable)
This network is utilized for all inter-server communications within the cloud infrastructure.
Recommended size: 255 IPs (CIDR /24)
Public Network (Publicly routable IP range)
This network is utilized for providing Public IP accessibility to the API endpoints within the cloud infrastructure.
inimum size: 8 IPs (CIDR /29)
VM Network (RFC1918 IP Range, not publicly routable)
This network is utilized for providing primary IP addresses to the cloud instances.
>Recommended size: 1024 IPs (CIDR /22)
Storage Network (RFC1918 IP Range, not publicly routable)
This network is utilized for all inter-server communications within the Object Storage infrastructure.
Recommended size: 255 IPs (CIDR /24)
Floating IP network (Publicly routable IP Range)
This network is utilized for providing Public IP accessibility to selected cloud instances.
Minimum size: 16 IPs (CIDR /28) OpenStack Software Specifications &Deployment
This section describes the software versions and combinations that work for the feature set included with this reference architecture.