Ask Your Question

Revision history [back]

I once had a conversation with a networking engineer from Rackspace, whose opinion was that OVS was overly complex and Linuxbridge was the right solution in most cases. That was in 2015 though, when we still used punched cards for programming.

Linuxbridge is simpler. Fewer moving parts, therefore less potential for problems and less hassle when something goes wrong. You need broader and deeper troubleshooting skills with OVS: With Linuxbridge tools like iptables and tcpdump suffice, whereas with OVS you need all of the above plus ovs-vsctl, ovs-ofctl and others. And try tracing packets that flow through an OVS bridge.

On the other hand, OVS seems to be the default, and the global knowledge base, including tutorials, documentation, bug reports etc is probably larger. Also, new networking functionality such as QoS a few years ago is often first implemented on OVS. Developers may find the richer feature set of OVS more comfortable, or perhaps they just go for the more popular solution first.

In terms of maturity, one could say that Linuxbridge has been around forever, whereas OVS is comparatively young. The Openstack Linuxbridge agent was also there first as far as I know. But OVS is used so much that it's probably as robust as Linuxbridge by now.