Hi Alfredo, thanks to your quick reply. 
I think my last statement is wrong, because when pfring_open is called for some 
physical interface, means that one ring will be created for all queues of this 
interface. when the packet rate reach some high value, overruns and drops 
happen.
As you mentioned, I should run multiple nprobe instances. So, I run ./pfcount 
-i eth1@0 for queue 0 of eth1.  It captured about 1/8 of all packets into this 
interface. for 82576, I have to create 8 nprobe instances for all 8 rx queues. 
That's too much for my box, cause I have 8 ports and that will be 64 instances. 
so is there any way to use one instance for more than one queues? I tried 
./pfcount -i multi:eth1@0eth1@1, but it doesn't work. any idea? thank you!


Alan






--------------------------------------------------------------------------------




------------------ 原始邮件 ------------------




 
 发件人: Alfredo Cardigliano  <[email protected]>
 

 
 发送时间: 2015-10-29 17:03:33
 

 
 收件人: ntop-misc <[email protected]> 
 

 
 抄送: (无) 
 

 
 主题: Re: [Ntop-misc] does nprobe support 82576 multiple channel?
 



Hi Alan
if you have an high traffic rate, you should use multiple nProbe instances, one 
per RSS channel.


Alfredo

On 29 Oct 2015, at 09:41, Wang <[email protected]> wrote:

Hi, guys. found that in nprobe 7,  "pfring_open" is called while not 
"pfring_open_multichannel", so this means that only one channel used in this 
situation. If this is true, for RSS netcard like 82576, will this cause some 
performance problem because only one queue works? 
 



Alan




_______________________________________________
Ntop-misc mailing list
[email protected]
http://listgateway.unipi.it/mailman/listinfo/ntop-misc





_______________________________________________
Ntop-misc mailing list
[email protected]
http://listgateway.unipi.it/mailman/listinfo/ntop-misc

Reply via email to