[ovs-dev] [RFC HSA 0/4] HSA for OVS prototype.

[Alex Wang]

Yeah, I think it will be useful.  I think current OVN's flow should be
simpler
and allow us to make more optimization/adjustment as OVN grows.

Thanks,
Alex Wang,

On Wed, May 27, 2015 at 9:56 AM, Ben Pfaff <blp at nicira.com> wrote:

> On Mon, Mar 30, 2015 at 03:46:25PM -0700, Alex Wang wrote:
> > This series implements a prototype of using Header Space Analysis (HSA)
> > for OVS OpenFlow table analysis.  The implementation allows users to find
> > all possible output ports with the input header format reachable from a
> > specified input port.  It also allows users to check if there is any loop
> > formed in the OpenFlow table.
> >
> > There are two major limitations of the prototype.  Firstly, only a
> limited
> > set of OFPACT_* actions are supported and extending to work with more
> actions
> > can be very complicated.  Secondly, the implementation is in single
> thread,
> > and can take 20+ mins to run on a production setup with ~100 OpenFlow
> rules.
> >
> > Since this is a very complicated work, I'd like to post the prototype out
> > for high-level discussions (e.g. feasibility, design, multi-threading).
>
> Hi Alex.
>
> I've delayed reviewing this mostly because I've been busy, but also
> because I wasn't too sure how useful it would be for most users.  But
> OVN might be a really good in-tree test case.  We could even incorporate
> it into the testsuite, if it is fast enough (and make it an optional
> part of the testsuite, if it is too slow).  Do you have an idea whether
> this HSA implementation would be appropriate for OVN?
>
> Thanks,
>
> Ben.
>