On Fri, Sep 6, 2019 at 1:14 AM Eelco Chaudron <[email protected]> wrote: > > > > On 5 Sep 2019, at 18:21, William Tu wrote: > > >> Did some more testing and with PVP I see a performance decrease, with > >> physical to physical I see an increase. > >> Tests are performed with a port redirect open flow rule on an ixgbe > >> (Xeon E5-2690 v4 2.60GHz): > >> > >> +-----------+-----------------+---------+---------+---------+---------+---------+---------+--------+ > >> | PVP | Number of flows | 64 | 128 | 256 | 512 > >> | > >> 768 | 1024 | 1514 | > >> +-----------+-----------------+---------+---------+---------+---------+---------+---------+--------+ > >> | master | 1000 | 737896 | 700643 | 682915 | 648386 > >> | > >> 621792 | 582821 | 527899 | > >> | Patch | 1000 | 734179 | 696515 | 676712 | 646536 > >> | > >> 619600 | 578546 | 519965 | > >> +-----------+-----------------+---------+---------+---------+---------+---------+---------+--------+ > >> > >> +-----------+-----------------+---------+---------+---------+---------+---------+---------+--------+ > >> | Port2Port | Number of flows | 64 | 128 | 256 | 512 > >> | > >> 768 | 1024 | 1514 | > >> +-----------+-----------------+---------+---------+---------+---------+---------+---------+--------+ > >> | master | 1000 | 3351114 | 3236581 | 3143710 | 2349598 > >> | > >> 1586276 | 1197304 | 814854 | > >> | Patch | 1000 | 3571733 | 3488697 | 3448908 | 2349593 > >> | > >> 1586276 | 1197304 | 814854 | > >> +-----------+-----------------+---------+---------+---------+---------+---------+---------+--------+ > > > > Hi Eleco, > > > > I'm wondering why you get only ~3.3Mpps for P2P test. Using ixgbe, I > > usually get at > > least 6Mpps using single flow (a single 64B UDP packet hitting single > > OpenFlow rule) > > Do you think it's due to have 1000 flows in your setup? > > I did a quick test and the number gets a bit better with fewer flows: > > "Physical loopback test, L3 flows[port redirect]" > ,Packet size > Number of flows,64 > 1,4111768 > 100,3904664 > 1000,3573564 > > I’m using a single of rule “in_port=eno1 actions=IN_PORT”, with a > single card, i.e. the same card in and out. > > > Another possible overhead is due to no rxhash in AF_XDP, so there is > > extra > > overhead doing flow_hash_5tuple and dp_packet_set_rss_hash > > But this should not influence the difference between our tests. Do you > use a single card/port in and out? I’m using a single queue and single > PMD thread.
I use single card with dual port (ex: eth2 and eth3), and single queue with single PMD for each netdev. BTW, the v2 patch about need_wakeup (I forgot to CC you)... https://patchwork.ozlabs.org/patch/1158678/ Thanks William _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
