netdev - Re: [ovs-dev] [PATCH net-next v1 1/1] net: openvswitch: ovs_packet_cmd_execute put sw_flow mainbody in stack

Open Source and information security mailing list archives

Message-ID: <OS3P286MB22951D8802775CC6DBF243F3F5A49@OS3P286MB2295.JPNP286.PROD.OUTLOOK.COM>
Date:   Mon, 20 Feb 2023 11:44:08 +0800
From:   Eddy Tao <taoyuan_eddy@...mail.com>
To:     Simon Horman <simon.horman@...igine.com>
Cc:     dev@...nvswitch.org, netdev@...r.kernel.org,
        linux-kernel@...r.kernel.org, Eric Dumazet <edumazet@...gle.com>,
        Jakub Kicinski <kuba@...nel.org>,
        Paolo Abeni <pabeni@...hat.com>,
        "David S. Miller" <davem@...emloft.net>
Subject: Re: [ovs-dev] [PATCH net-next v1 1/1] net: openvswitch:
 ovs_packet_cmd_execute put sw_flow mainbody in stack

More explanation to the meaning of performance data

 >>Mode Iterations   Variance    Average

Iterations: the number of executions of the same test case (each 
iteration get a performance value of perf[n] )

Average:  sum of all execution and then divided by 'n_iterations'. Below 
is the pseudo code

                for (sum=0,i=0;i<iterations;i++) sum+= perf[i];

                average = sum/iterations

Variance: A percentage value describing the overall deviation of 
performance results from average. Below is the pseudo code

                for (deviation=0,i=0;i<iterations;i++) deviation+= 
square(perf[i] - average);

                variance = ((square_root(deviation/iterations)) * 
100)/average

On 2023/2/20 11:04, Eddy Tao wrote:
> Hi， Simon:
>
>     To have better visibility of the effect of the patch, i did 
> another test below
>
> Disabling data-path flow installation to steer traffic to slow path 
> only, thus I can observe the performance on slow path, where 
> ovs_packet_cmd_execute is extensively used
>
>
> Testing topology
>
>             |-----|
>       nic1--|     |--nic1
>       nic2--|     |--nic2
> VM1(16cpus) | ovs |   VM2(16 cpus)
>       nic3--|     |--nic3
>       nic4--|     |--nic4
>             |-----|
> 2 netperf client threads on each vnic
>
> netperf -H $peer -p $((port+$i)) -t TCP_STREAM  -l 60
> netperf -H $peer -p $((port+$i)) -t TCP_RR  -l 60 -- -R 1 -r 120,240
> netperf -H $peer -p $((port+$i)) -t TCP_CRR -l 60 -- -R 1 -r 120,240
>
>   Mode Iterations   Variance    Average
>
> TCP_STREAM     10      %3.83       1433 ==> before the change
> TCP_STREAM     10      %3.39       1504 ==> after  the change
>
> TCP_RR         10      %2.35      45145 ==> before the change
> TCP_RR         10      %1.06      47250 ==> after  the change
>
> TCP_CRR        10      %0.54      11310 ==> before the change
> TCP_CRR        10      %2.64      12741 ==> after  the change
>
>
> Considering the size and simplicity of the patch, i would say the 
> performance benefit is decent.
>
> Thanks
>
> eddy
>
>
> _______________________________________________
> dev mailing list
> dev@...nvswitch.org
> https://mail.openvswitch.org/mailman/listinfo/ovs-dev

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives