[ovs-dev] [PATCH v2] Simplify kernel sFlow implementation

[Ben Pfaff]

On Thu, Aug 18, 2011 at 10:43:17PM -0700, Neil McKee wrote:
> 
> On Aug 18, 2011, at 6:55 PM, Jesse Gross wrote:
> 
> > On Aug 19, 2011 6:41 AM, "Neil McKee" <neil.mckee at inmon.com> wrote:
> > 
> > > Another check worth confirming is that the random number generator
> > > you are using will always converge exactly to the intended 1-in-N
> > > mean (before too long).  That gives you another fall-back
> > > position.
> > 
> > I'm not sure what you mean by "fall-back position"?
> > 
> I meant that you have the option of just incrementing the sample-pool
> by the mean (N) for every sample taken.

That's a clever idea.

For the record, that's probably not a good approach for Open vSwitch
because in various overload situation the buffers between kernel and
userspace can overflow.  When that happens, packets are lost and there
is no easy way to count how many packets were lost, so we would lose
track of the sample pool at that point.