On 25.01.2012 14:18, Nikolay Denev wrote:
On Jan 25, 2012, at 1:59 PM, Ivan Voras wrote:
On 24/01/2012 17:53, Marcin Markowski wrote:
On 24.01.2012 14:22, Ivan Voras wrote:
On Mon, Jan 23, 2012 at 12:20 PM, Marcin
Markowski<mmarkow...@leon.pl>wrote:
(on 9.0 we can see also kernel thread named {ix0 que} using 100%
CPU),
hw.ixgbe.num_queues=16
If there really are 16 hardware queues, shouldn't there be 16
kernel
threads for queue processing?
There are 16 threads, but only one of them consumes 100% CPU and
the others
do not use more than 5% CPU:
http://pastebin.com/BWDWh8kW
You need Jack to confirm it but this looks like a serious problem /
bottleneck. It just shouldn't be like that (if the test is exactly the
same).
Is this by any chance non-IP traffic (or encapsulated in some way).
I remember similar thread :
http://freebsd.1045724.n5.nabble.com/igb-ixgbe-RSS-RX-queues-for-non-IP-traffic-td4778961.html
Most of the traffic received by the interface is a TCP/UDP, but there
is also a PPPoE (~2.5Gbps/s).
The whole is divided into a large number of VLANs (>500)
--
Marcin Markowski
_______________________________________________
freebsd-performance@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-performance
To unsubscribe, send any mail to "freebsd-performance-unsubscr...@freebsd.org"
