[dpdk-dev] KNI performance numbers...

Maciej Grochowski maciej.grochowski at codilime.com
Thu Jun 25 13:20:13 CEST 2015


I meet similar issue with KNI connected VM, but In my case I run 2 VM
guests based on KNI and measure network performance between them:

sesion:

### I just started demo with kni

./build/kni -c 0xf0 -n 4 -- -P -p 0x3 --config="(0,4,6,8),(1,5,7,9)"

###starting...

###set kni on vEthX to connect (as in example)

echo 1 > /sys/class/net/vEth0_0/sock_en
fd=`cat /sys/class/net/vEth0_0/sock_fd`

## start first guest VM
kvm -nographic -name vm1 -cpu host -m 2048 -smp 1 -hda
.../debian_squeeze_amd64.qcow2 -netdev tap,fd=$fd,id=hostnet1,vhost=on
-device virtio-net-pci,netdev=hostnet1,id=net1,bus=pci.0,addr=0x4

## start second guest VM
echo 1 > /sys/class/net/vEth1_0/sock_en
fd=`cat /sys/class/net/vEth1_0/sock_fd`

kvm -nographic -name vm2 -cpu host -m 2048 -smp 1 -hda
.../debian_squeeze2_amd64.qcow2 -netdev tap,fd=$fd,id=hostnet1,vhost=on
-device virtio-net-pci,netdev=hostnet1,id=net1,bus=pci.0,addr=0x4

###END: ustawiam 2 kvm z virtual guestem


### first VM node start server
 netserver -p 22113

### performance from second VM guest to first (server) using netperf

root at debian-amd64:~# netperf -H 10.0.0.200 -p 22113 -t TCP_STREAM
MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to
10.0.0.200 () port 0 AF_INET
Recv   Send    Send
Socket Socket  Message  Elapsed
Size   Size    Size     Time     Throughput
bytes  bytes   bytes    secs.    10^6bits/sec
 87380  16384  16384    10.01     219.86

So I got 220M between two VM using KNI, but it was only experiment (I
didn't analyze it deeply)

On Wed, Jun 24, 2015 at 7:58 AM, Vithal S Mohare <vmohare at arubanetworks.com>
wrote:

> Hi,
>
> I am running DPDP KNI application on linux (3.18 kernel) VM (ESXi 5.5),
> directly connected to another linux box to measure throughput using  iperf
> tool.  Link speed: 1Gbps.   Maximum throughput I get is 50% with 1470
> Bytes.  With 512B pkt sizes, throughput drops to 282 Mbps.
>
> Tried using KNI loopback modes (and traffic from Ixia), but no change in
> throughput.
>
> KNI is running in single thread mode.  One lcore for rx, one for tx and
> another fir kni thread.
>
> Is the result expected?  Has anybody got better numbers?  Appreciate for
> input and relevant info.
>
> Thanks,
> -Vithal
>


More information about the dev mailing list