rsday, January 12, 2017 5:40 PM
> To: Ilya Maximets ; Bodireddy, Bhanuprakash
> ; Aaron Conole
> Cc: d...@openvswitch.org; Daniele Di Proietto
> Subject: Re: [ovs-dev] [PATCH] netdev-dpdk: Use intermediate queue during
> packet transmission.
>
> Hi Ilya,
> thanks for your
12:47 PM
> To: Bodireddy, Bhanuprakash ; Aaron
> Conole
> Cc: d...@openvswitch.org; Daniele Di Proietto ;
> Thadeu Lima de Souza Cascardo ; Fischetti, Antonio
> ; Markus Magnusson
>
> Subject: Re: [ovs-dev] [PATCH] netdev-dpdk: Use intermediate queue during
> packet
Daniele Di Proietto ;
>> Thadeu Lima de Souza Cascardo ; Fischetti, Antonio
>> ; Markus Magnusson
>>
>> Subject: Re: [ovs-dev] [PATCH] netdev-dpdk: Use intermediate queue during
>> packet transmission.
>>
>> On 19.12.2016 21:05, Bodireddy, Bhanuprakas
o
>; Markus Magnusson
>
>Subject: Re: [ovs-dev] [PATCH] netdev-dpdk: Use intermediate queue during
>packet transmission.
>
>On 19.12.2016 21:05, Bodireddy, Bhanuprakash wrote:
>> Thanks Ilya and Aaron for reviewing this patch and providing your
>comments, my reply inline.
>>
, 2016 8:41 AM
>> To: Aaron Conole ; Bodireddy, Bhanuprakash
>>
>> Cc: d...@openvswitch.org; Daniele Di Proietto ;
>> Thadeu Lima de Souza Cascardo ; Fischetti, Antonio
>> ; Markus Magnusson
>>
>> Subject: Re: [ovs-dev] [PATCH] netdev-dpdk: Use intermediate qu
witch.org; Daniele Di Proietto ;
>Thadeu Lima de Souza Cascardo ; Fischetti, Antonio
>; Markus Magnusson
>
>Subject: Re: [ovs-dev] [PATCH] netdev-dpdk: Use intermediate queue during
>packet transmission.
>
>Hi,
>
>I didn't test this patch yet. Bhanu, could you ple
Hi,
I didn't test this patch yet. Bhanu, could you please describe
your test scenario and performance results in more details.
It'll be nice if you provide throughput and latency measurement
results for different scenarios and packet sizes. Latency is
important here.
About the patch itself:
1.
Hi Bhanu,
Bhanuprakash Bodireddy writes:
> In exact match cache processing on an EMC hit, packets are queued in to
> batches matching the flow. Thereafter, packets are processed in batches
> for faster packet processing. This particularly is inefficient if there
> are fewer packets in a batch as
In exact match cache processing on an EMC hit, packets are queued in to
batches matching the flow. Thereafter, packets are processed in batches
for faster packet processing. This particularly is inefficient if there
are fewer packets in a batch as rte_eth_tx_burst() incurs expensive MMIO
write.
Th