From: Clarylin L <clearasu@gmail.com>
To: "Tan, Jianfeng" <jianfeng.tan@intel.com>
Cc: dev@dpdk.org
Subject: Re: [dpdk-dev] L3 Forwarding performance of DPDK on virtio
Date: Wed, 20 Jan 2016 19:18:42 -0800 [thread overview]
Message-ID: <CAKj5Mjs=6wbT44RnmfqS1Z=wqV41SxS=e8VcQJgKNvxzPy4pCw@mail.gmail.com> (raw)
In-Reply-To: <56A03376.2030903@intel.com>
Sorry. It's L2 forwarding.
I used testpmd with forwarding mode, like
testpmd --pci-blacklist 0000:00:05.0 -c f -n 4 -- --portmask 3 -i
--total-num-mbufs=20000 --nb-cores=3 --mbcache=512 --burst=512
--forward-mode=mac --eth-peer=0,90:e2:ba:9f:95:94
--eth-peer=1,90:e2:ba:9f:95:95
On Wed, Jan 20, 2016 at 5:25 PM, Tan, Jianfeng <jianfeng.tan@intel.com>
wrote:
>
> Hello!
>
>
> On 1/21/2016 7:51 AM, Clarylin L wrote:
>
>> I am running dpdk within a virtual guest as a L3 forwarder.
>>
>>
>> The VM has two ports connecting to two linux bridges (in turn connecting
>> two physical ports). DPDK is used to forward between these two ports (one
>> port connected to traffic generator and the other connected to sink). I
>> used iperf to test the throughput.
>>
>>
>> If the VM/DPDK is running on passthrough, it can achieve around 10G
>> end-to-end (from traffic generator to sink) throughput. However if the
>> VM/DPDK is running on virtio (virtio-net-pmd), it achieves just 150M
>> throughput, which is a huge degrade.
>>
>>
>> On the virtio, I also measured the throughput between the traffic
>> generator
>> and its connected port on VM, as well as throughput between the sink and
>> it's VM port. Both legs show around 7.5G throughput. So I guess forwarding
>> within the VM (from one port to the other) would be a big killer of the
>> performance.
>>
>>
>> Any suggestion on how I can root cause the poor performance issue, or any
>> idea on performance tuning techniques for virtio? thanks a lot!
>>
>
> The L3 forwarder, you mentioned, is the l3fwd example in DPDK? If so, I
> doubt it can work well with virtio, see another thread "Add API to get
> packet type info".
>
> Thanks,
> Jianfeng
>
prev parent reply other threads:[~2016-01-21 3:18 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-01-20 23:51 Clarylin L
2016-01-21 1:25 ` Tan, Jianfeng
2016-01-21 3:18 ` Clarylin L [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='CAKj5Mjs=6wbT44RnmfqS1Z=wqV41SxS=e8VcQJgKNvxzPy4pCw@mail.gmail.com' \
--to=clearasu@gmail.com \
--cc=dev@dpdk.org \
--cc=jianfeng.tan@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).