Hi all, Just wanted to comment on what I sent a couple of weeks back:
> On Nov 3, 2016, at 7:47 PM, Luis Gomez <ece...@gmail.com> wrote: > > Hi all, > > Few comments after looking at the OpenFlow CI test results: > > 1) As commented this morning Carbon shows more stability than Boron in the > cbench and longevity tests, so please do not forget to cherry-pick changes > Carbon->Boron. Longevity is now good in both Boron + Carbon so I closed https://bugs.opendaylight.org/show_bug.cgi?id=5271 <https://bugs.opendaylight.org/show_bug.cgi?id=5271>. > > 2) After looking at Cbench test in detail: > > - Current Throughput test running in CI does not seem to represent controller > performance (Latency test is still fine). This is because I get much better > results when I run the same test locally so I think public cloud networking > is probably impacting here. > - When testing Cbench throughput locally, Boron/Carbon is 2x faster than > Beryllium. This for example cannot be appreciated in the CI test. > > 3) I will open bug for the 2 things I observe in the new scalability tests > and I also commented this morning: > > - Switch connect time is much faster in Beryllium than in Boron/Carbon. The > switch connect time also seems to increase substantially the more switches > are connected to controller. > - Max number of links are double in Beryllium than in Boron/Carbon. The > failure I see is controller cannot see some of the switches after some > iterations. After more debugging, it seems the differences between Beryllium vs Boron/Carbon are due to different mininet/OVS version used in the test. Unfortunately Beryllium does not work well with OVS 2.5 so it cannot be updated to match the Boron/Carbon test. > > BR/Luis >
_______________________________________________ openflowplugin-dev mailing list openflowplugin-dev@lists.opendaylight.org https://lists.opendaylight.org/mailman/listinfo/openflowplugin-dev
