Open Stack

On Wed, Dec 3, 2014 at 3:32 AM, Abhijeet Rastogi
<abhijeet.1989 at gmail.com> wrote:
> Hi Antonio,
>
> Thanks for the link. Could you explain a bit about "We want to avoid a
> single point of failure, so we decided to give direct access to this
> network from the compute node."?
>
> What exactly did you mean by that?

Basically we would like to avoid passing through a network node when
it's not needed, to avoid a potential bottleneck + SPoF.

With Juno there is also the Distributed Routing (DVR) option if you
are using Floating IPs, but we haven't tested it yet, and still has
some disadvantages.

In the setup described, the router used by the VMs is an external
router (actually 2 routers in VRRP), the traffic from the VM goes
directly to the router, and there is no encapsulation involved. This
should increase speed an reliability.

.a.

-- 
antonio.s.messina at gmail.com
antonio.messina at uzh.ch                     +41 (0)44 635 42 22
S3IT: Service and Support for Science IT   http://www.s3it.uzh.ch/
University of Zurich
Winterthurerstrasse 190
CH-8057 Zurich Switzerland