[dpdk-dev] Is VFIO driver's performance better than IGB_UIO?

[dpdk-dev] Is VFIO driver's performance better than IGB_UIO?

[Linhaifeng]

I have test the VFIO driver and IGB_UIO driver by l2fwd for many times. I find that the VFIO driver’s performance is not better than the IGB_UIO.

Is something wrong with my test? My test is as follow:

1.       bind two 82599 ether to VFIO  ./tools/dpdk_nic_bind.py -b vfio-pci 03:00.0 03:00.1

2.       run the l2fwd to watch the stats info.

3.       Bind the 82599 ehter to IGB_UIO  ./tools/dpdk_nic_bind.py -b igb_uio 03:00.0 03:00.1

4.       run the l2fwd to watch the stats info.




The result of test is :（Mpps）
VFIO-64-PHY-PHY

IGBUIO-64-PHY-PHY

VFIO-512-PHY-PHY

IGBUIO-512-PHY-PHY

2.6235456

2.3467793

1.9432854

1.9432753

2.5724822

2.5405128

1.9432777

1.9432832

2.2418318

2.5154781

1.9395376

1.9432291

2.470847

2.551112

1.9432767

1.9432756

2.5092176

2.4965851

1.9432705

1.9432733

2.51572

2.4703292

1.9432637

1.9432669

2.6293656

2.472452

1.9432719

1.943283

2.480364

2.6295004

1.9432775

1.943286

2.4242182

2.613162

1.943268

1.9432663



I have two questions:

1.       If the result is wrong how can I test the VFIO and IGB_UIO

2.       If the result is normal why use VFIO?

Re: [dpdk-dev] Is VFIO driver's performance better than IGB_UIO?

[Vincent JARDIN]

On 08/08/2014 09:41, Linhaifeng wrote:
> I have test the VFIO driver and IGB_UIO driver by l2fwd for many times. I find that the VFIO driver’s performance is not better than the IGB_UIO.

You are right, under some conditions UIO is faster, VFIO provides
safety. The best solution is a PMD without UIO, neither VFIO.

which CPU are you using?

Best regards,
  Vincent

[dpdk-dev] 答复: Is VFIO driver's performance better than IGB_UIO?

[Linhaifeng]

Thank you very much.

My cpu is "Intel(R) Xeon(R) CPU           E5620  @ 2.40GHz" 


-----邮件原件-----
发件人: Vincent JARDIN [mailto:vincent.jardin@6wind.com] 
发送时间: 2014年8月8日 15:46
收件人: Linhaifeng
抄送: dev@dpdk.org; lixiao (H); Guofeng (E)
主题: Re: [dpdk-dev] Is VFIO driver's performance better than IGB_UIO?

On 08/08/2014 09:41, Linhaifeng wrote:
> I have test the VFIO driver and IGB_UIO driver by l2fwd for many times. I find that the VFIO driver’s performance is not better than the IGB_UIO.

You are right, under some conditions UIO is faster, VFIO provides safety. The best solution is a PMD without UIO, neither VFIO.

which CPU are you using?

Best regards,
  Vincent

[dpdk-dev] Is VFIO driver's performance better than IGB_UIO?

[Vincent JARDIN]

> My cpu is "Intel(R) Xeon(R) CPU           E5620  @ 2.40GHz"

It is a Westmere if I am correct. So,use UIO.

Re: [dpdk-dev] Is VFIO driver's performance better than IGB_UIO?

[Chris Wright]

* Vincent JARDIN (vincent.jardin@6wind.com) wrote:
> > My cpu is "Intel(R) Xeon(R) CPU           E5620  @ 2.40GHz"
> 
> It is a Westmere if I am correct. So,use UIO.

My experience is that the CPU isn't as an important part of the platform
as the chipset.  IOW, VFIO requires full IOMMU isolation, and the IOMMU
cost depends on the chipset.

Vincent, is that what you were implying by cpu, or do you see something
else?

thanks,
-chris