[Ntop-misc] PF_RING DNA Testing: Throttling? or Tweaking

[Alfredo Cardigliano]

Charles
if you load the driver with single queue (RSS=0,0) you just need to use pfcount -i dna0.
Instead, if you use more queues, you can:
- run more pfcount instances with pfcount -i dna0 at X (where X is the queue)
- or run pfcount_multichannel

Regards
Alfredo

On Apr 27, 2012, at 11:30 AM, Charles H. Leggett wrote:

> Alfredo et al.,
> 
> Thanks, I was able to load the ixgbe DNA driver with the modified load_dna_driver.sh (with RSS=0,0,0,0) script but, now …
> 
> I am not sure how to call the interface with pfcount or tcpdump or whatever to avoid capturing only "from the first one ("-i dna0" == "-i dna0 at 0")".
> 
> I am trying this:
> 
> [chl at sensor1 src]$ sudo pfcount -i dna0
> 
> How should I call the interface in order to capture all of the queues?
> 
> Thanks for your help on this.
> 
> On Apr 25, 2012, at 3:29 PM, Alfredo Cardigliano wrote:
>> Charles
>> it looks like you are loading the dna driver with 64 queues, but you are capturing just from the first one ("-i dna0" == "-i dna0 at 0").
>> Please try loading the driver with (edit the load_dna_driver.sh script):
>> insmod ./ixgbe.ko RSS=0,0,0,0
>> (mono queue)
>> 
>> Regards
>> Alfredo
>> 
>> On Apr 25, 2012, at 7:12 PM, Charles H. Leggett wrote:
>> 
>>> PF_RING'ers,
>>> 
>>> I am testing PF_RING with both the "pf_ring aware in transparent_mode=2" and the unlicensed DNA ixgbe driver.
>>> 
>>> With pfcount, I am seeing very different results (see "tests" below).
>>> 
>>> In a nutshell while not losing any packets, I am losing 86% of the throughput.
>>> 
>>> I figure that either the testing driver is throttling in addition to limiting captures to 5 mins (I don't believe this is the case) or, I have to do some serious tweaking to get it to use more cores, etc.
>>> 
>>> I have included info from /proc on pf_ring at the bottom of test 2.
>>> 
>>> Please have a look and see if you can point me in the right direction.
>>> 
>>> Thanks.
>>> 
>>> Tests:
>>> 
>>> 1) "pf_ring aware in transparent_mode=2" ixgbe:
>>> 
>>> [chl at sensor1 ~]$ sudo pfcount -i p6p1
>>> Using PF_RING v.5.3.0
>>> Capturing from p6p1 [90:E2:BA:07:B9:68]
>>> # Device RX channels: 16
>>> # Polling threads:    1
>>> =========================
>>> Absolute Stats: [125770 pkts rcvd][0 pkts dropped]
>>> Total Pkts=125770/Dropped=0.0 %
>>> 125'770 pkts - 87'542'885 bytes
>>> =========================
>>> 
>>> =========================
>>> Absolute Stats: [417424 pkts rcvd][0 pkts dropped]
>>> Total Pkts=417424/Dropped=0.0 %
>>> 417'424 pkts - 328'852'997 bytes [417'375.16 pkt/sec - 2'630.51 Mbit/sec]
>>> =========================
>>> Actual Stats: 291654 pkts [1'000.11 ms][291'619.88 pps/1.93 Gbps]
>>> =========================
>>> 
>>> 2) unlicensed DNA ixgbe:
>>> 
>>> [chl at sensor1 ~]$ sudo pfcount -i dna0
>>> ###################################################
>>> # ERROR: You do not seem to have a valid DNA license for dna0 [Intel 10 Gbit ixgbe 82599-based].
>>> # ERROR: Please get one at http://shop.ntop.org/.
>>> ###################################################
>>> # We're now working in demo mode with packet capture
>>> # and transmission limited to 5 minutes
>>> ###################################################
>>> Using PF_RING v.5.3.0
>>> Capturing from dna0 [90:E2:BA:07:B9:68]
>>> # Device RX channels: 64
>>> # Polling threads:    1
>>> =========================
>>> Absolute Stats: [20012 pkts rcvd][0 pkts dropped]
>>> Total Pkts=20012/Dropped=0.0 %
>>> 20'012 pkts - 16'791'945 bytes
>>> =========================
>>> 
>>> =========================
>>> Absolute Stats: [38783 pkts rcvd][0 pkts dropped]
>>> Total Pkts=38783/Dropped=0.0 %
>>> 38'783 pkts - 31'467'381 bytes [38'770.32 pkt/sec - 251.66 Mbit/sec]
>>> =========================
>>> Actual Stats: 18771 pkts [1'000.32 ms][18'764.88 pps/0.12 Gbps]
>>> =========================
>>> 
>>> [chl at sensor1 src]$ cat /proc/net/pf_ring/info 
>>> PF_RING Version     : 5.3.0 ($Revision: exported$)
>>> Ring slots          : 4096
>>> Slot version        : 13
>>> Capture TX          : Yes [RX+TX]
>>> IP Defragment       : No
>>> Socket Mode         : Standard
>>> Transparent mode    : No (mode 2)
>>> Total rings         : 0
>>> Total plugins       : 0
>>> 
>>> [chl at sensor1 src]$ cat /proc/net/pf_ring/dev/dna0/info 
>>> Name:              dna0
>>> Index:             18
>>> Address:           90:E2:BA:07:B9:68
>>> Polling Mode:      DNA
>>> Type:              Ethernet
>>> Family:            Intel ixgbe 82599
>>> Max # TX Queues:   64
>>> # Used RX Queues:  64
>>> 
>>> --
>>> Charles H. Leggett
>>> Information Security Architect
>>> chl at uga.edu
>>> 
>>> 
>>> 
>>> 
>>> 
>>> 
>>> _______________________________________________
>>> Ntop-misc mailing list
>>> Ntop-misc at listgateway.unipi.it
>>> http://listgateway.unipi.it/mailman/listinfo/ntop-misc
>> 
>> _______________________________________________
>> Ntop-misc mailing list
>> Ntop-misc at listgateway.unipi.it
>> http://listgateway.unipi.it/mailman/listinfo/ntop-misc
>> 
> 
> --
> Charles H. Leggett
> Information Security Architect
> chl at uga.edu
> 
> 
> 
> 
>