[ovs-dev] How stable is Distributed Switch?

Justin Pettit jpettit at nicira.com
Sun Aug 30 08:15:58 UTC 2009


> I wish to know, is Open-V-Switch stable enough for using with VMs  
> which span across different physical hosts?
> I was comparing vSphere and OpenVSwitch.

I'm not sure what you mean by "stable enough".  The code is quite  
stable at this point and is already being used by some external  
users.  It will be going through a more proper beta release soon.  We  
use it internally.

Is your question actually about how it works as a distributed switch?   
Open vSwitch has full support for OpenFlow (http:// 
openflowswitch.org), which allows a centralized controller remote  
access to its forwarding tables.  For configuring parts of the switch  
that are not covered by OpenFlow (e.g., traffic policing, bond  
configuration, NetFlow, etc.), we've defined a management protocol  
that is similar OpenFlow.  However, the OpenFlow Consortium (of which  
many of the Open vSwitch developers are a part) is in the process of  
defining an official management protocol, which we will adopt when its  
design is finalized.  All this means that some sort of controller is  
needed to allow policy and configuration state to follow VMs as they  
migrate.

> Either OpenVSwitch feature list is too abstract or it has really  
> long way to go.
> It will be great if someone could add the lacking-features to the To- 
> Do list.


I think your feedback may be a bit too abstract.  ;-)  What specific  
features do you think are lacking?

--Justin






More information about the dev mailing list