Hello, Alfredo! Thank you for answer!
My case is very interesting. With hardware NIC capabilities (5-tuple filters) we can direct specific traffic to certain NIC queue with enabled PF_RING traffic processor. All another traffic could be processed with Linux stack. It's zero overhead solution. Direct and not so intelligent solution to this problem look like this. We will process all traffic with PF_RING and flow back non interested for us traffic back to Linux network stack. It's not straightforward to implement, has non zero overhead and bug-aware. With ability to open specific NIC queue we could combine monitoring task and production network service on same server and same NIC. Now I should use two separate servers for this task. Thank you! On Wed, Mar 25, 2015 at 7:36 PM, Alfredo Cardigliano <[email protected]> wrote: > Hi Pavel > when you open a queue pf_ring now detaches the whole interface from kernel > for coherency, > I guess you agree that in common use cases this is preferable to avoid time > windows where half of the traffic is sent to the kernel. > We should probably try to support also your use case somehow, as a special > case. > > Alfredo > >> On 25 Mar 2015, at 17:00, Pavel Odintsov <[email protected]> wrote: >> >> Hello, folks! >> >> I need open only one NIC hardware queue with PF_RING ZC. Another 7 >> queues should be handled by Linux kernel for normal traffic. >> >> Is it possible with PF_RING/ZC? >> >> -- >> 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
