On Tue, Jun 18, 2013 at 04:52:20PM +0000, Rajahalme, Jarno (NSN - FI/Espoo) wrote: > > On Jun 18, 2013, at 0:18 , ext Ben Pfaff wrote: > > > On Mon, Jun 17, 2013 at 11:45:49AM +0000, Rajahalme, Jarno (NSN - FI/Espoo) > > wrote: > >> On Jun 17, 2013, at 3:55 , Jarno Rajahalme wrote: > >> > >>> The simplest way for an OpenFlow controller to refer to a (set of) flows > >>> is by a controller-issued flow cookie. Make this fast by inserting > >>> flows with non-zero cookies to a hmap, and use that when flows are > >>> queried with a full cookie mask. > >>> > >> > >> Here are some measurements on this, using a flow table with > >> 138260 different flow like below, each flow with a different IP prefix, > >> a priority corresponding to the prefix length, and a unique cookie: > > > > The measurements are pretty awesome, but I don't like abusing an hmap > > to store data that might have duplicates (the fact that you had to > > exclude elements with a zero cookie value shows that you likely > > thought about this too). > > I did not even stop to think about possibly abusing the hmap, sorry ;-)
I've wanted a multimap structure for a long time now, and your patch gave me an excuse to design and build one, so I'm happy ;-) > > So I built another data structure that doesn't have this problem. > > Will you review it? > > http://openvswitch.org/pipermail/dev/2013-June/028556.html > > I changed to use the new hindex instead, and run the measurements on > full-masked cookies again. Unsurprisingly the results were the same :-) I'm not surprised either, but it gives some confidence that the hindex code is correct. Are you happy with the hindex patch? _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev