Good to know ZC works as expected

Luca

> On 12 Dec 2014, at 17:33, Pavel Odintsov <[email protected]> wrote:
> 
> Yep, I opened zc:ethX and everything works fine. But ksoftirqd still
> overload my system because packets flow into Linux Network Stack too
> :(
> 
> On Fri, Dec 12, 2014 at 7:29 PM, Luca Deri <[email protected]> wrote:
>> ok but for using zc you need to open “zc:eth4” whereas “eth4” uses sthe
>> standard PF_RING
>> 
>> Luca
>> 
>> On 12 Dec 2014, at 17:23, Pavel Odintsov <[email protected]> wrote:
>> 
>> Maybe it will be relevant and I have 2 NIC's with 2 ports each in this
>> server:
>> 
>> 0a:00.0 Ethernet controller: Intel Corporation 82599EB 10-Gigabit
>> SFI/SFP+ Network Connection (rev 01)
>> 0a:00.1 Ethernet controller: Intel Corporation 82599EB 10-Gigabit
>> SFI/SFP+ Network Connection (rev 01)
>> 0d:00.0 Ethernet controller: Intel Corporation 82599EB 10-Gigabit
>> SFI/SFP+ Network Connection (rev 01)
>> 0d:00.1 Ethernet controller: Intel Corporation 82599EB 10-Gigabit
>> SFI/SFP+ Network Connection (rev 01)
>> 
>> 
>> 
>> On Fri, Dec 12, 2014 at 7:16 PM, Pavel Odintsov
>> <[email protected]> wrote:
>> 
>> Oh, very strange!
>> 
>> In transparent_mode=2 quick_mode=1 I still can see packets at
>> interface with standard pcap tcpdump:
>> tcpdump -i eth4 -c 10
>> tcpdump: WARNING: eth4: no IPv4 address assigned
>> tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
>> listening on eth4, link-type EN10MB (Ethernet), capture size 65535 bytes
>> 19:15:19.916730 IP 10.10.10.100.49645 > 10.10.10.200.0: UDP, length 0
>> 19:15:19.916733 IP 10.10.10.100.41742 > 10.10.10.200.0: UDP, length 0
>> 19:15:19.916734 IP 10.10.10.100.30047 > 10.10.10.200.0: UDP, length 0
>> 19:15:19.916735 IP 10.10.10.100.41743 > 10.10.10.200.0: UDP, length 0
>> 19:15:19.916736 IP 10.10.10.100.38649 > 10.10.10.200.0: UDP, length 0
>> 19:15:19.916737 IP 10.10.10.100.30048 > 10.10.10.200.0: UDP, length 0
>> 19:15:19.916739 IP 10.10.10.100.49646 > 10.10.10.200.0: UDP, length 0
>> 19:15:19.916740 IP 10.10.10.100.52632 > 10.10.10.200.0: UDP, length 0
>> 19:15:19.916741 IP 10.10.10.100.41744 > 10.10.10.200.0: UDP, length 0
>> 19:15:19.916742 IP 10.10.10.100.30049 > 10.10.10.200.0: UDP, length 0
>> 10 packets captured
>> 2980 packets received by filter
>> 2839 packets dropped by kernel
>> 
>> On Fri, Dec 12, 2014 at 7:11 PM, Pavel Odintsov
>> <[email protected]> wrote:
>> 
>> Hello, Luca!
>> 
>> I'm loading ixgbe patched ixgbe driver, loaded pf_ring, start hping
>> flood from another machine and got 90% of kernel load from ksoftirqd.
>> I did not started any toolkit yet :(
>> 
>> But ZC looks like working because it work many times faster then
>> vanilla PF_RING and process 2 mpps of packets! And it complains about
>> license:
>> 
>> # ERROR: You do not seem to have a valid PF_RING ZC license
>> 6.0.2.140923 for eth4 [Intel 10 Gbit ixgbe 82599-based]
>> # ERROR: Please get one at http://shop.ntop.org/.
>> 
>> 
>> On Fri, Dec 12, 2014 at 7:02 PM, Luca Deri <[email protected]> wrote:
>> 
>> Pavel
>> if you have ksoftirq up, I believe you are not using ZC. Please explain what
>> you are doing?
>> 
>> 
>> Luca
>> 
>> On 12 Dec 2014, at 16:24, Pavel Odintsov <[email protected]> wrote:
>> 
>> Total number of generated packets per second flood was about 1million
>> packets per second, CPU E5-2407 0 @ 2.20GHz.
>> 
>> On Fri, Dec 12, 2014 at 6:19 PM, Pavel Odintsov
>> <[email protected]> wrote:
>> 
>> Hello!
>> 
>> I'm using PF_RING 6.0.2 with Debian 7 Wheezy 3.2.0-4-amd64. I
>> installed ixgbe driver with ZC support and loaded it with
>> load_driver.sh.
>> 
>> I don't need any processing and linux stack thus I set: transparent_mode=2.
>> 
>> After this I run udp flood from 6 hping instances from another server:
>> hping3 -I eth1 --udp --flood 10.10.10.200
>> 
>> After this I run top and got this:
>> %Cpu(s):  0.0 us,  0.0 sy,  0.0 ni, 90.3 id,  0.0 wa,  0.0 hi,  9.7 si,  0.0
>> st
>>  3 root      20   0     0    0    0 S  78.5  0.0   5:40.02
>> ksoftirqd/0
>> htop looks like this:
>> 1  [||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||93.9%]
>>  Tasks: 24, 3 thr, 49 kthr; 1 running
>> 2  [
>> 0.0%]     Load average: 0.05 0.20 0.22
>> 3  [
>> 0.0%]     Uptime: 18 days, 22:04:44
>> 4  [                                                             0.0%]
>> Mem[|||||||||||||||                                      2661/32207MB]
>> Swp[                                                         0/8190MB]
>> 
>> 
>> I even tried to enable quick_mode=1 but ksoftirqd consuming almost
>> whole CPU core.
>> 
>> Thank you for your attention!
>> 
>> --
>> Sincerely yours, Pavel Odintsov
>> 
>> 
>> 
>> 
>> --
>> Sincerely yours, Pavel Odintsov
>> _______________________________________________
>> 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
>> 
>> 
>> 
>> 
>> --
>> Sincerely yours, Pavel Odintsov
>> 
>> 
>> 
>> 
>> --
>> Sincerely yours, Pavel Odintsov
>> 
>> 
>> 
>> 
>> --
>> Sincerely yours, Pavel Odintsov
>> _______________________________________________
>> 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
> 
> 
> 
> -- 
> Sincerely yours, Pavel Odintsov
> _______________________________________________
> 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