startup.conf is
unix {
nodaemon
cli-listen localhost:5002
log /tmp/vpp.log
full-coredump
}
api-trace {
on
}
cpu{
main-core 0
corelist-worker 1
}
dpdk{
uio-driver igb_uio
dev 0000:01:00.0
dev 0000:02:00.0
}
2016-10-21 15:03 GMT+08:00 yu scott <[email protected]>:
> Q1:
>
> I run vpp in kvm vm,and the software nic frontend is virtio backend is
> vhost-user.
> I test the vpp route performance with Tester(vpp code version is
> stable/16.09）, vm connect Tester with dpdk
> ovs datapath
>
>
> vpp startup.conf is
>
>
> vppctl config is
>
>
> set int state interface1 up
> set int state interface2 up
> set int ip address interface1 172.16.1.1/24
> set int ip address interface2 172.16.2.1/24
>
> kvm config is as :
> https://wiki.fd.io/view/VPP/How_To_Optimize_Performance_(System_Tuning)
>
> vpp route forward is 80kpps. My host use E5 2630 v3 CPU
>
> In my mind ,vpp is high performance,so I think my test config for vpp may be
> not suitable for
> vpp high performance. Is my config right for vpp high performance ?
> Does some config make vpp faster I don't use ?
>
>
> Q2:
>
> After performance test，I exec the cli "show error"
> it show <Tx packet drop (dpdk tx faiulre)> item has so many packet.
>
> In my experience，If
> vpp have no cpu to forward packets, it will lose packet in dpdk rx, not tx.
>
> If lose packet in tx, I think the reason is tx nic backend in
> ovs dpdk datapath not
> timely get the packet away. But in my test environment ,dpdk ovs datapath
> get packet rate is 10mpps，vpp forward rate is 80kpps.
>
> Hope to you reply :-)
>
> Best Regards
>
> Yu
>
>
_______________________________________________
vpp-dev mailing list
[email protected]
https://lists.fd.io/mailman/listinfo/vpp-dev
