Hi Antonio,

I think I need to test another platform. It may give us more information.
I will choose one arm board to have a try when I'm free.

Tony

Antonio Quartulli <a...@unstable.cc> 于2024年1月30日周二 19:02写道:
>
> Hi,
>
> On 29/01/2024 05:25, Tony He wrote:
> > Hi Antonio,
> >
> > I'm using ovpn-dco which is backported to v4.14 based on your latest
> > code. My topology is:
> >
> > LAN PC -- openwrt router running openvpn server -- WAN PC running openvpn 
> > client
> > Router is with two mips64 cores.
> >
> > I use the iperf3 to test speed between LAN PC and WAN PC.
> > The result is sometime the performance is good(~280Mbps) while
> > sometimes is bad(~140Mbps).
> >   When the performance is bad, one of two CPUs is 100% idle. When the
> > performance is good, two CPUs are busy. However, I don't see the issue
> > when ipsec is tested in the same test env. Two cores are always used
> > for ipsec. So , can ovpn-dco use all cpu cores to get max performance?
>
> ovpn-dco uses more than core in order to perform different operations,
> but more parallelism on traffic processing can definitely be implemented
> (patches are welcome ;)).
>
> Now it's hard to tell if what you are seeing is the result of this
> implementation detail or something else, especially because in some
> cases you get higher throughput.
>
>
> Cheers,
>
> --
> Antonio Quartulli


_______________________________________________
Openvpn-devel mailing list
Openvpn-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/openvpn-devel

Reply via email to