Date: Fri, 28 Feb 2020 22:39:58 +0300 From: Slawa Olhovchenkov <slw@zxy.spb.ru> To: Vincenzo Maffione <vmaffione@freebsd.org> Cc: FreeBSD Net <freebsd-net@freebsd.org> Subject: Re: Intel NETMAP performance and packet type Message-ID: <20200228193958.GR8012@zxy.spb.ru> In-Reply-To: <CA%2B_eA9iZE9=fFFhOTo7mo3DN5mWErcP1dpDrf4_tLgAoqVPdEw@mail.gmail.com> References: <20200203204447.GD8028@zxy.spb.ru> <CA%2B_eA9it5iB7ERiDx0b7zN54atPZz64wW75nWeZrqTD58SWyLQ@mail.gmail.com> <20200225150924.GM8012@zxy.spb.ru> <CA%2B_eA9hqsUj7bXnq0%2BprqsxTr7_f9WE-5%2BJwUA==Xy0PvXAG=Q@mail.gmail.com> <20200227201650.GO8012@zxy.spb.ru> <CA%2B_eA9iZE9=fFFhOTo7mo3DN5mWErcP1dpDrf4_tLgAoqVPdEw@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, Feb 28, 2020 at 06:31:36PM +0100, Vincenzo Maffione wrote: > Il giorno gio 27 feb 2020 alle ore 21:17 Slawa Olhovchenkov <slw@zxy.spb.ru> > ha scritto: > > > On Thu, Feb 27, 2020 at 06:51:54PM +0100, Vincenzo Maffione wrote: > > > > > Hi, > > > So, the issue is not the payload. > > > If you look at the avg_batch statistics reported by pkt-gen, you'll see > > > that in the ACK-flood experiment you have 4.92, whereas in the SYN-flood > > > case you have 17.5. The batch is the number of packets (well, actually > > > netmap descriptors, but in this case it's the same) that you receive (or > > > transmit) for each poll() invocation. > > > So in the first case you end up doing much more poll() calls, hence the > > > higher per-packet overhead and the lower packet-rate. > > > > > > Why is the poll() called more frequently? That depends on packet timing > > and > > > interrupt rate. There must be something different on your packet > > generator > > > that produces this effect (e.g. different burstiness, or maybe the packet > > > generator is not able to saturate the 10G link)? > > > > No, I am capture netstat output -- raw packet rate is the same. > > Also, I am change card to chelsio T5 and don't see issuse. > > > > This is payload issuse, at driver level. > > > > That's not possible, since netmap does not even look into the payload. Netmap don't look, NIC can look. I mean driver don't coretly/filly switch off all acceleration and ACK flood caused badly handled event in Intel driver. Don't see for chelsio. > Can you please report the per-queue interrupt rate in both cases (ACK-flood > and SYN-flood)? > You can use something like `vmstat -i -w1 | grep ix` to monitor the > interrupt rate. > Or probably you can also use `sysctl -a dev.ix | grep interrupt_rate` > > > > > In any case, I would suggest measuring the RX interrupt rate, and check > > > that it's higher in the ACK-flood case. Then you can try to lower the > > > interrupt rate by tuning the interrupt moderation features of the Intel > > NIC > > > (e,g. limit hw.ix.max_interrupt_rate and disable hw.ix.enable_aim or > > > similar). > > > By playing with the interrupt moderation you should be able to increase > > the > > > avg_batch, and then increase throghput. > > > > Already limited. > > > > Limited to which value? Have you tried to decrease max_interrupt_rate even > more? 16000, don't try different values for this case.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20200228193958.GR8012>