> On 16 Dec 2014, at 01:49, Jesse Bowling <[email protected]> wrote:
> 
> 
> On Dec 12, 2014, at 5:55 PM, Alfredo Cardigliano <[email protected]> wrote:
> 
>> 
>>> On 12 Dec 2014, at 20:44, Jesse Bowling <[email protected]> wrote:
>>> 
>>> In the vanilla PF_RING world, one could look under /proc/net/pf_ring and 
>>> find files for each process using PF_RING, and these files contained 
>>> information on packets seen, dropped, processed, as well as memory 
>>> information. I’d like to know if there are similar things when using the ZC 
>>> drivers.
>> 
>> With ZC kernel is bypassed, this means the application reads stats directly 
>> from the NIC, and usually print them under /proc/net/pf_ring/stats
> 
> OK, perfect!
> 
>> 
>>> I did find that when using the zbalance_ipc example program to bind two 
>>> interfaces together and then deliver that information to a single process 
>>> there was a file under /proc/net/pf_ring/stats for the process, but it did 
>>> not see to contain useful information in terms of performance.
>>> 
>>> What is the recommended way to obtain statistics on ZC performance (or 
>>> really, app performance using ZC)?
>> 
>> /proc is the way, please let us know what is missing under /stats that you 
>> think is useful
> 
> I suppose I got used to the way that PF_RING would let you know how the app 
> was performing; in other words, if my snort instances were consuming too 
> slowly and packets in the ring were being overwritten before being read, I 
> could see that in the “drop” counts...Or perhaps the issue that was really 
> being reported was that PF_RING was dropping the packets before they could be 
> inserted in the ring? Either way, I liked being able to get a sense of my 
> apps performance via the PF_RING stats. It seems like the new stats don’t 
> really give you that info…

If your snort instances are consuming too slowly, incoming packets are being 
discarded. Snort is able to read stats about discarded packets through the 
DAQ-ZC.

Alfredo

>> 
>>> I notice zbalance_ipc does print statistics to the screen while running; 
>>> can these be disabled?
>> 
>> There is no parameter at the moment, but they are not printed in daemon mode.
> 
> Ah! I was missing the daemon mode flag...
>> 
>>> Would you expect zbalance_ipc to be used in a similar fashion as 
>>> pfdnaclustermaster was for DNA? 
>> 
>> Yes, they are supposed to be equivalent, do think zbalance_ipc is missing 
>> useful features available in pfdnaclustermaster?
> 
> Really just getting started with this, and wanted to make sure I was on the 
> correct track!
> 
> Thank you so much for your replies!
> 
> Cheers,
> 
> Jesse
> _______________________________________________
> 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