Cloud Computing 9 min read

Choosing the Right Network Component for OpenStack: Nova-Network vs Neutron

An in-depth overview explains OpenStack’s Nova-Network and Neutron architectures, compares their networking models, discusses node roles, high-availability strategies, overlay technologies like VxLAN, and offers practical guidance for selecting the right network component based on business scenarios.

Efficient Ops
Efficient Ops
Efficient Ops
Choosing the Right Network Component for OpenStack: Nova-Network vs Neutron

Sharing Content

Before answering the question, a brief introduction to

Nova-Network

and

Neutron

is provided.

Nova-Network

Nova-Network is the networking solution used in OpenStack versions prior to Folsom, as a submodule of the Compute project (Nova). It supports Flat Network, Flat DHCP Network, VLAN Network, using technologies such as linux-bridge and VLAN.

Neutron

Neutron, originally named Quantum, became a separate networking service starting with the Folsom release and was renamed in Havana due to trademark issues. It supports Local, VLAN, GRE, VXLAN networks and utilizes linux-bridge, OVS, VXLAN, GRE, VLAN, OpenFlow, etc. Neutron is essentially the evolution of Nova-Network.

Nova-Network’s Flat and FlatDHCP modes share a common IP pool, differing only in allocation method, and provide no isolation, suitable for small, static topologies where security groups can be used.

VLAN mode requires device configuration and is suitable for modest isolation needs with limited scale.

Neutron Architecture

Control node : runs the Neutron server, maintaining network information in the database.

Network node : runs the L3 agent; in DVR mode, the compute node also runs L3 and L3 NAT agents.

Compute node : may run plugin agents depending on configuration.

In early Neutron versions without DVR, both east‑west and north‑south traffic had to pass through the network node, creating a bottleneck. DVR reduces this load but the network node can still be a point of failure.

High‑availability solutions for the network node include:

Replacing the network node with integrated hardware and plugins.

Master‑slave mode using VRRP and conntrack synchronization (still experimental).

Merging network functions into compute nodes, moving NAT to them (requires significant redesign).

Leveraging SDN/OpenFlow to handle NAT, load balancing, and routing directly in the data plane, eliminating the need for a dedicated network node.

Overlay Networks

Overlay technologies such as VxLAN, NVGRE, GRE, Geneve, and STT are widely used. VxLAN, carried over UDP, is the most common today, with hardware offload support emerging in modern NICs.

Choosing the appropriate OpenStack networking component depends on the specific business scenario; newer Neutron features are attractive but must be thoroughly tested before production use.

Interactive Q&A

Which method does JD.com use? Currently VLAN is most suitable for private clouds.

Can traffic be directly steered to physical switches?

Which backend storage solutions (e.g., GlusterFS, Ceph) are production‑ready?

What JD.com business workloads run on its cloud platform?

SDNCloud NetworkingOpenStackNeutronNova-Network
Efficient Ops
Written by

Efficient Ops

This public account is maintained by Xiaotianguo and friends, regularly publishing widely-read original technical articles. We focus on operations transformation and accompany you throughout your operations career, growing together happily.

0 followers
Reader feedback

How this landed with the community

login Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.