Eric W. Biederman wrote:
>>When an outgoing packet has the loopback destination addres, the
>>skbuff is filled with the network namespace. So the loopback packets
>>never go outside the namespace. This approach facilitate the migration
>>of loopback because identification is done by network namespace and
>>not by address. The loopback has been benchmarked by tbench and the
>>overhead is roughly 1.5 %
>
>
> Ugh. 1.5% is noticeable.
We will see with all private network namespace ...
>
> I think it is cheaper to have one loopback device per namespace.
> Which removes the need for a skbuff tag.
OpenVZ Forum
Members
Search
Help
Register
Login
Home
