On Mon, Aug 27, 2012 at 03:45:28PM +0530, Dan Cross wrote: > On Mon, Aug 27, 2012 at 3:33 PM, Francois Tigeot <ftig...@wolfpond.org> wrote: > > On Mon, Aug 27, 2012 at 11:42:00AM +0200, Jan Lentfer wrote: > >> > >> Can you run the same tests with pgbench running on a second box? > > > > I gave it a try; DragonFly performance increased linearly with the number > > of clients, without any relation to actual hardware resources. > > What did the rest look like? In particular, if one cuts of the > possibility of the Scientific Linux "loopback cuts out the IP stack" > optimiziation hypothesis, how does its performance change relative to > the others?
I only did a few points with DragonFly-3.1 for lack of time; I can't answer this question with the data I have currently. -- Francois Tigeot
