[Snort-devel] Large Packet Drop with SNort-2.9.80 as compared to Snort-2.9.7.6

Dheeraj Gupta dheeraj.gupta4 at ...2499...
Thu Dec 17 01:16:52 EST 2015


Hi,

The test was run for the same PCAP so number of packets is same in both
cases (9220233). The packet I/O totals as output by two snorts are:

Snort-2.9.8.0
------------------------
===============================================================================
Run time for packet processing was 783.512468 seconds
Snort processed 9220233 packets.
Snort ran for 0 days 0 hours 13 minutes 3 seconds
   Pkts/min:       709248
   Pkts/sec:        11775
===============================================================================
===============================================================================
Packet I/O Totals:
   Received:      9220233
   Analyzed:      9220233 (100.000%)
    Dropped:            0 (  0.000%)
   Filtered:            0 (  0.000%)
Outstanding:            0 (  0.000%)
   Injected:            0
===============================================================================


Snort-2.9.7.6
-----------------------

===============================================================================
Run time for packet processing was 547.131014 seconds
Snort processed 9220233 packets.
Snort ran for 0 days 0 hours 9 minutes 7 seconds
   Pkts/min:      1024470
   Pkts/sec:        16856
===============================================================================
===============================================================================
Packet I/O Totals:
   Received:      9220233
   Analyzed:      9220233 (100.000%)
    Dropped:            0 (  0.000%)
   Filtered:            0 (  0.000%)
Outstanding:            0 (  0.000%)
   Injected:            0
===============================================================================

Again as the test is against a static PCAP, there will be no drops.
However, in this test Snort-2.9.8.0 is almost 30% slower (processes about
11.7K pkts/s as against 16.8K pkts/s) than Snort-2.9.7.6. When used with
live traffic, wouldn't this cause increased packet drops?

Regards,
Dheeraj

On Wed, Dec 16, 2015 at 8:02 PM, Nageswara Rao A.V.K (navk) <navk at ...3482....>
wrote:

> You did not provide “Packet I/O Totals:” for this test.
>
> We have to compare that data.
>
>
>
> I don’t think previous stats will applicable here.
>
> Because the number of pkts are different here.
>
>
>
> Best Regards,
>
> -ANR
>
>
>
> *From:* Dheeraj Gupta [mailto:dheeraj.gupta4 at ...2499...]
> *Sent:* Wednesday, December 16, 2015 5:16 PM
> *To:* Nageswara Rao A.V.K (navk)
> *Cc:* snort-devel at lists.sourceforge.net
> *Subject:* Re: [Snort-devel] Large Packet Drop with SNort-2.9.80 as
> compared to Snort-2.9.7.6
>
>
>
> Hi,
>
> I captured a large PCAP (6.6G ~9M packets) and analyzed it through both
> Snort-2.9.7.6 and 2.9.8.0 with almost identical configuration file (memcap
> etc.). Since SO rules for Snort-2.9.7.6 cannot be used with 2.9.8.0, so
> number of rules for 2.9.8.0 was less (about 11k) as compared to 2.9.7.6
> (12k).
>
> Here is a summary of end of run stats
>
> Snort-2.9.7.6
>
>
> ===============================================================================
> Run time for packet processing was 547.131014 seconds
> Snort processed 9220233 packets.
> Snort ran for 0 days 0 hours 9 minutes 7 seconds
>    Pkts/min:      1024470
>    Pkts/sec:        16856
>
> ===============================================================================
>
> Snort-2.9.8.0
>
> ===============================================================================
> Run time for packet processing was 783.512468 seconds
> Snort processed 9220233 packets.
> Snort ran for 0 days 0 hours 13 minutes 3 seconds
>    Pkts/min:       709248
>    Pkts/sec:        11775
>
> ===============================================================================
>
> snort.conf is attached
>
>
>
> On Tue, Dec 15, 2015 at 10:03 AM, Dheeraj Gupta <dheeraj.gupta4 at ...3054....>
> wrote:
>
> Hi,
>
> The traffic is captured from a live interface, so it is not exactly same.
> However, it is from the same network and same network filter over a
> contiguous time range. So, characteristics of the trafic are broadly the
> same i.e. most of it is user browsing data. The reason I wrote this e-mail
> is because on a weekday, we have an average 100-150 Mbps on the wire and
> Snort-2.9.7.6 reported less losses (<10%). However, Snort-2.9.8.0 reported
> over 40% drops with comparable traffic load/pattern.
>
> Snort logs do not have any additional entry apart from session pruned due
> to timeout/stale (same in both cases).
>
> Regards,
>
> Dheeraj
>
>
>
> On Tue, Dec 15, 2015 at 8:43 AM, Nageswara Rao A.V.K (navk) <
> navk at ...3461...> wrote:
>
> Hi Dheeraj,
>
>    We need more info to get in to conclusion.
>
>
>
> Are you passing same traffic in both scenario’s??
>
>
>
> Did you verify snort logs ??
>
> You may know the reason for pkt drops.
>
>
>
> We did not notice this problems in our observation.
>
> More details may help us to analyze the problem.
>
>
>
> Best Regards,
>
> -ANR
>
>
>
> *From:* Dheeraj Gupta [mailto:dheeraj.gupta4 at ...2499...]
> *Sent:* Monday, December 14, 2015 11:30 AM
> *To:* snort-devel at lists.sourceforge.net
> *Subject:* [Snort-devel] Large Packet Drop with SNort-2.9.80 as compared
> to Snort-2.9.7.6
>
>
>
> Hi,
>
> I just upgraded to Snort-2.9.8.0 from Snort-2.9.7.6. Before the upgrade
> one of my sensors showed (somewhat expected) packet drops. However, after
> the upgrade the packet drop increased significantly even though the number
> of rules decreased (as SO rules are not in use with 2.9.8.0). I am still
> using Snort-2.9.7.6 rulesets (as advised by you).
>
> Here is a snip from my snort.stats file for 2.9.8.0
>
> #time,pkt_drop_percent,wire_mbits_per_sec.realtime
> 1450068900,33.873,124.415
> 1450069200,23.718,121.253
> 1450069500,26.014,120.349
> 1450069800,26.368,120.821
> 1450070100,23.706,116.493
> 1450070400,21.039,121.363
>
> For Snort-2.9.7.6, the snip is
> #time,pkt_drop_percent,wire_mbits_per_sec.realtime
> 1450071180,0.000,79.159
> 1450071480,0.000,118.671
> 1450071780,2.146,132.186
> 1450072080,8.337,130.408
>
>
>
> Looking at end-of-snort stats. This is for 2.9.8.0
>
> Packet I/O Totals:
>    Received:    804563792
>    Analyzed:    388361098 ( 48.270%)
>     Dropped:    298207658 ( 27.042%)
>    Filtered:    415840607 ( 51.685%)
>    Outstanding:       362087 (  0.045%)
>    Injected:            0
>
> And this is for 2.9.7.6
>
> Packet I/O Totals:
>    Received:     60969886
>    Analyzed:     30035104 ( 49.262%)
>     Dropped:       742645 (  1.203%)
>    Filtered:     30927585 ( 50.726%)
>    Outstanding:         7197 (  0.012%)
>    Injected:            0
>
> I have a longish BPF filter, so is the filtered count an indication of the
> amount of traffic which was filtered by that filter?
>
> Also is dropped count a subset of analyzed count or received count? I ask
> this because it appears
>
> received_count = analyzed + filtered
>
> so dropped_count doesn't really fit in
>
>
>
> Regards,
>
> Dheeraj
>
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.snort.org/pipermail/snort-devel/attachments/20151217/2e5fc22c/attachment.html>


More information about the Snort-devel mailing list