[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CANn89iJh4_urk440UAWYxr0RvVLjBJ+d4B-hDovSJ+d_0oWfMw@mail.gmail.com>
Date: Thu, 31 May 2018 10:05:10 -0400
From: Eric Dumazet <edumazet@...gle.com>
To: moshe@...lanox.com
Cc: Tariq Toukan <tariqt@...lanox.com>,
Eric Dumazet <eric.dumazet@...il.com>, aring@...atatu.com,
David Miller <davem@...emloft.net>,
netdev <netdev@...r.kernel.org>, Florian Westphal <fw@...len.de>,
Herbert Xu <herbert@...dor.apana.org.au>,
Thomas Graf <tgraf@...g.ch>,
Jesper Dangaard Brouer <brouer@...hat.com>,
Alexander Aring <alex.aring@...il.com>,
Stefan Schmidt <stefan@....samsung.com>,
Kirill Tkhai <ktkhai@...tuozzo.com>,
Eran Ben Elisha <eranbe@...lanox.com>
Subject: Re: [PATCH v4 net-next 00/19] inet: frags: bring rhashtables to IP defrag
On Thu, May 31, 2018 at 8:18 AM Moshe Shemesh <moshe@...lanox.com> wrote:
>
>
>
> I do see big improvement after changing the 3 parameters as Eric suggested:
> /proc/sys/net/ipv6/ip6frag_time set to 2
> /proc/sys/net/ipv6/ip6frag_low_thresh set to 104857600
> /proc/sys/net/ipv6/ip6frag_high_thresh set to 78643200
>
>
> [root@...-l-vrt-67100-104 linux-stable]# netperf -H
> fe80::7efe:90ff:fed5:bb48%ens9,inet6 -t udp_stream --
> MIGRATED UDP STREAM TEST from ::0 (::) port 0 AF_INET6 to
> fe80::7efe:90ff:fed5:bb48%ens9 () port 0 AF_INET6
> Socket Message Elapsed Messages
> Size Size Time Okay Errors Throughput
> bytes bytes secs # # 10^6bits/sec
>
> 212992 65507 10.00 156387 0 8194.60
> 212992 10.00 76901 4029.57
>
> #kernel
> Ip6InReceives 7107999 0.0
> Ip6InDelivers 114126 0.0
> Ip6OutRequests 47 0.0
> Ip6ReasmTimeout 5115 0.0
> Ip6ReasmReqds 7107987 0.0
> Ip6ReasmOKs 114114 0.0
> Ip6ReasmFails 1714146 0.0
> ...
> Udp6InDatagrams 112486 0.0
> Udp6InErrors 1629 0.0
> Udp6RcvbufErrors 1629 0.0
> ...
>
> While before these parameters settings I got:
> [root@...-l-vrt-67100-104 ~]# netperf -H
> fe80::e61d:2dff:feca:c7c3%ens9,inet6 -t udp_stream --
> MIGRATED UDP STREAM TEST from ::0 (::) port 0 AF_INET6 to
> fe80::e61d:2dff:feca:c7c3%ens9 () port 0 AF_INET6
> Socket Message Elapsed Messages
> Size Size Time Okay Errors Throughput
> bytes bytes secs # # 10^6bits/sec
>
> 212992 65507 10.00 145419 0 7620.35
> 212992 10.00 285 14.93
>
> #kernel
> Ip6InReceives 6665965 0.0
> Ip6InDelivers 300 0.0
> Ip6OutRequests 9 0.0
> Ip6ReasmReqds 6665950 0.0
> Ip6ReasmOKs 285 0.0
> Ip6ReasmFails 6650890 0.0
> ...
> Udp6InDatagrams 286 0.0
>
>
> however, before the patchset, I got much better results:
> [root@...-l-vrt-67100-104 linux-stable]# netperf -H
> fe80::7efe:90ff:fed5:bb48%ens9,inet6 -t udp_stream --
> MIGRATED UDP STREAM TEST from ::0 (::) port 0 AF_INET6 to
> fe80::7efe:90ff:fed5:bb48%ens9 () port 0 AF_INET6
> Socket Message Elapsed Messages
> Size Size Time Okay Errors Throughput
> bytes bytes secs # # 10^6bits/sec
>
> 212992 65507 10.00 158935 0 8328.32
> 212992 10.00 144652 7579.88
>
>
> #kernel
> Ip6InReceives 7088903 0.0
> Ip6InDelivers 154117 0.0
> Ip6OutRequests 9 0.0
> Ip6ReasmReqds 7088889 0.0
> Ip6ReasmOKs 154103 0.0
> ...
> Udp6InDatagrams 144653 0.0
> Udp6InErrors 9451 0.0
> Udp6RcvbufErrors 9451 0.0
>
>
Hi Moshe
Your environment seems to be very lossy.
Frags and packets losses do not mix well, there is nothing really
magic we can do with that,
unless we can reserve GB of memory for frags.
Which is exactly was the reason for my patches in the first place.
Thanks.
Powered by blists - more mailing lists