[Snort-users] Snort + PF_RING + DAQ

livio Ricciulli livio at ...15149...
Tue Sep 4 20:00:01 EDT 2012


On 09/04/2012 04:12 PM, Luca Deri wrote:
> Hi Livio,
> if you have a look at http://www.ntop.org/wp-content/uploads/2012/09/Snort_over_DNA_Silicom_30_07_2012_1.pdf you will see that the speed bump of DNA with respect to non-DNA PF_RING DAQ ranges from +20% to over +1400%. So I won't say "slightly better performance".
>
> Regards Luca
Slight in the message below was referring to the difference between 24 
snort processes (freely scheduled by Linux using NAPI) and 16 snort 
processes using DNA (DNA wins).
Thanks for the write-up!
It kind of confirms our numbers if I understand them right (almost 
linear speedup 14/16):
The 1400% measures the improvement of single thread af_packet  (4 
threads x 4 machines) to mutiple thread NAPI (16 threads x 4 machines)?
And 20% compares multiple thread NAPI (16 threads x 4 machines) to 
multiple-thread DNA (16 threads x 4 machines)?
Right?

With our workload (large EDU trace) the bottleneck seems to be more on 
the CPU and less IO. When IO starts being more of a bottleneck, then DNA 
clearly wins big (I would not call 20% slight).

Livio.
> On Sep 5, 2012, at 12:54 AM, livio Ricciulli<livio at ...15149...>  wrote:
>
>>> CPU Binding is something important, QUEUE wise if you bind a snort
>>> process to the same network QUEUE
>>> then you can clearly start to benchmark. If you spread the network
>>> queue load on multiple CPU and do not bind process
>>> to the same CPU then your adding context switching in the mix which i
>>> think is bad at high throuput.
>> In pfring lingo this is called DNA and does give slightly better performance
>> which supports your claim:
>> see https://www.metaflows.com/technology/10-gbps-pf_ring-2/
>>
>> We found, though, that with NAPI and letting the Linux scheduler loose
>> on 24 threads
>> works just as well but gives you much better flexibility (you can have
>> multiple
>> applications share the same interface for example which you cannot do
>> with DNA).
>>
>> So, your theory is correct but it does not make a big enough difference,
>> (on our appliances).
>> And I doubt it would solve Peter's problem. But again, it is hard to
>> generalize and I might be wrong..
>>
>> Livio.
>>
>>
>> ------------------------------------------------------------------------------
>> Live Security Virtual Conference
>> Exclusive live event will cover all the ways today's security and
>> threat landscape has changed and how IT managers can respond. Discussions
>> will include endpoint security, mobile security and the latest in malware
>> threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
>> _______________________________________________
>> Snort-users mailing list
>> Snort-users at lists.sourceforge.net
>> Go to this URL to change user options or unsubscribe:
>> https://lists.sourceforge.net/lists/listinfo/snort-users
>> Snort-users list archive:
>> http://www.geocrawler.com/redir-sf.php3?list=snort-users
>>
>> Please visit http://blog.snort.org to stay current on all the latest Snort news!





More information about the Snort-users mailing list