Re: [ovs-discuss] ovs-dpdk performance is not good

Re: [ovs-discuss] ovs-dpdk performance is not good

[Traynor, Kevin]

[-- Attachment #1.1.1: Type: text/plain, Size: 1700 bytes --]

(re-adding the ovs-discuss list)

This might be better on the dpdk dev mailing list. For the OVS part, see this thread http://openvswitch.org/pipermail/discuss/2015-July/018095.html

Kevin.

From: Na Zhu [mailto:zhunatuzi@gmail.com]
Sent: Wednesday, July 15, 2015 6:16 AM
To: Traynor, Kevin
Subject: Re: [ovs-discuss] ovs-dpdk performance is not good

Hi Kevin,

The interface MTU is 1500, the TCP message size is 16384 and the UDP message size is 65507.
How to use DPDK virtio PMD?

2015-07-14 20:25 GMT+08:00 Traynor, Kevin <kevin.traynor@intel.com<mailto:kevin.traynor@intel.com>>:

From: discuss [mailto:discuss-bounces@openvswitch.org<mailto:discuss-bounces@openvswitch.org>] On Behalf Of Na Zhu
Sent: Monday, July 13, 2015 3:15 AM
To: bugs@openvswitch.org<mailto:bugs@openvswitch.org>
Subject: [ovs-discuss] ovs-dpdk performance is not good

Dear all,

I want to use ovs-dpdk to improve my nfv performance. But when i compare the throughput between standard ovs and ovs-dpdk, the ovs is better, does anyone know why?

I use netperf to test the throughput.
use vhost-net to test standard ovs.
use vhost-user to test ovs-dpdk.
My topology is as follow:

[内嵌图片 1]

The result is that standard ovs performance is better. Throughput unit Mbps.
[内嵌图片 2]

[内嵌图片 3]


[kt] I would check your core affinitization to ensure that the vswitchd
pmd is on a separate core to the vCPUs (set with other_config:pmd-cpu-mask).
Also, this test is not using the DPDK vitrio PMD in the guest which provides
performance gains.

What packet sizes are you using? you should see a greater gain from DPDK
at lower packet sizes (i.e. more PPS)




[-- Attachment #1.1.2: Type: text/html, Size: 10093 bytes --]

[-- Attachment #1.2: image001.png --]
[-- Type: image/png, Size: 16645 bytes --]

[-- Attachment #1.3: image002.png --]
[-- Type: image/png, Size: 11447 bytes --]

[-- Attachment #1.4: image003.png --]
[-- Type: image/png, Size: 9044 bytes --]

[-- Attachment #2: Type: text/plain, Size: 141 bytes --]

_______________________________________________
discuss mailing list
discuss@openvswitch.org
http://openvswitch.org/mailman/listinfo/discuss

Re: [ovs-discuss] ovs-dpdk performance is not good

[Ouyang, Changchun]

On 7/16/2015 9:45 PM, Traynor, Kevin wrote:
>
> (re-adding the ovs-discuss list)
>
> This might be better on the dpdk dev mailing list. For the OVS part, 
> see this thread 
> http://openvswitch.org/pipermail/discuss/2015-July/018095.html
>
> Kevin.
>
> *From:*Na Zhu [mailto:zhunatuzi@gmail.com]
> *Sent:* Wednesday, July 15, 2015 6:16 AM
> *To:* Traynor, Kevin
> *Subject:* Re: [ovs-discuss] ovs-dpdk performance is not good
>
> Hi Kevin,
>
> The interface MTU is 1500, the TCP message size is 16384 and the UDP 
> message size is 65507.
>
> How to use DPDK virtio PMD?
>
in DPDK virtio PMD, it uses mergeable feature to support jumbo frame,
the mergeable feature need negotiate with vhost on the backend,
so if ovs enable the mergeable feature, and virtio can succeed in 
negotiating this feature,
then jumbo frame can be supported.

thanks
Changchun

> 2015-07-14 20:25 GMT+08:00 Traynor, Kevin <kevin.traynor@intel.com 
> <mailto:kevin.traynor@intel.com>>:
>
>     *From:*discuss [mailto:discuss-bounces@openvswitch.org
>     <mailto:discuss-bounces@openvswitch.org>] *On Behalf Of *Na Zhu
>     *Sent:* Monday, July 13, 2015 3:15 AM
>     *To:* bugs@openvswitch.org <mailto:bugs@openvswitch.org>
>     *Subject:* [ovs-discuss] ovs-dpdk performance is not good
>
>     Dear all,
>
>     I want to use ovs-dpdk to improve my nfv performance. But when i
>     compare the throughput between standard ovs and ovs-dpdk, the ovs
>     is better, does anyone know why?
>
>     I use netperf to test the throughput.
>
>     use vhost-net to test standard ovs.
>
>     use vhost-user to test ovs-dpdk.
>
>     My topology is as follow:
>
>     内嵌图片 1
>
>     The result is that standard ovs performance is better. Throughput
>     unit Mbps.
>
>     内嵌图片 2
>
>     内嵌图片 3
>
>     [kt] I would check your core affinitization to ensure that the
>     vswitchd
>
>     pmd is on a separate core to the vCPUs (set with
>     other_config:pmd-cpu-mask).
>
>     Also, this test is not using the DPDK vitrio PMD in the guest
>     which provides
>
>     performance gains.
>
>     What packet sizes are you using? you should see a greater gain
>     from DPDK
>
>     at lower packet sizes (i.e. more PPS)
>

Re: [ovs-discuss] ovs-dpdk performance is not good

[Na Zhu]

[-- Attachment #1.1.1: Type: text/plain, Size: 2587 bytes --]

Hi,

Can you tell me how to enable the mergeable feature in ovs?

BR

2015-07-17 11:04 GMT+08:00 Ouyang, Changchun <changchun.ouyang-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>:

>
>
> On 7/16/2015 9:45 PM, Traynor, Kevin wrote:
>
>  (re-adding the ovs-discuss list)
>
>
>
> This might be better on the dpdk dev mailing list. For the OVS part, see
> this thread http://openvswitch.org/pipermail/discuss/2015-July/018095.html
>
>
>
> Kevin.
>
>
>
> *From:* Na Zhu [mailto:zhunatuzi-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org <zhunatuzi-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>]
> *Sent:* Wednesday, July 15, 2015 6:16 AM
> *To:* Traynor, Kevin
> *Subject:* Re: [ovs-discuss] ovs-dpdk performance is not good
>
>
>
> Hi Kevin,
>
>
>
> The interface MTU is 1500, the TCP message size is 16384 and the UDP
> message size is 65507.
>
> How to use DPDK virtio PMD?
>
> in DPDK virtio PMD, it uses mergeable feature to support jumbo frame,
> the mergeable feature need negotiate with vhost on the backend,
> so if ovs enable the mergeable feature, and virtio can succeed in
> negotiating this feature,
> then jumbo frame can be supported.
>
> thanks
> Changchun
>
>
>
> 2015-07-14 20:25 GMT+08:00 Traynor, Kevin <kevin.traynor-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>:
>
>
>
> *From:* discuss [mailto:discuss-bounces-yBygre7rU0TnMu66kgdUjQ@public.gmane.org] *On Behalf Of *Na
> Zhu
> *Sent:* Monday, July 13, 2015 3:15 AM
> *To:* bugs-yBygre7rU0TnMu66kgdUjQ@public.gmane.org
> *Subject:* [ovs-discuss] ovs-dpdk performance is not good
>
>
>
> Dear all,
>
>
>
> I want to use ovs-dpdk to improve my nfv performance. But when i compare
> the throughput between standard ovs and ovs-dpdk, the ovs is better, does
> anyone know why?
>
>
>
> I use netperf to test the throughput.
>
> use vhost-net to test standard ovs.
>
> use vhost-user to test ovs-dpdk.
>
> My topology is as follow:
>
>
>
> [image: 内嵌图片 1]
>
>
>
> The result is that standard ovs performance is better. Throughput unit
> Mbps.
>
> [image: 内嵌图片 2]
>
>
>
> [image: 内嵌图片 3]
>
>
>
>
>
> [kt] I would check your core affinitization to ensure that the vswitchd
>
> pmd is on a separate core to the vCPUs (set with
> other_config:pmd-cpu-mask).
>
> Also, this test is not using the DPDK vitrio PMD in the guest which
> provides
>
> performance gains.
>
>
>
> What packet sizes are you using? you should see a greater gain from DPDK
>
> at lower packet sizes (i.e. more PPS)
>
>
>
>
>
>
>
>
>

[-- Attachment #1.1.2: Type: text/html, Size: 11883 bytes --]

[-- Attachment #1.2: Type: image/png, Size: 11447 bytes --]

[-- Attachment #1.3: Type: image/png, Size: 9044 bytes --]

[-- Attachment #1.4: Type: image/png, Size: 16645 bytes --]

[-- Attachment #2: Type: text/plain, Size: 141 bytes --]

_______________________________________________
discuss mailing list
discuss@openvswitch.org
http://openvswitch.org/mailman/listinfo/discuss