On 20.02.2015 12:38, Kalil de A. Carvalho wrote: > Hello all. > > We are moving from Xenserver 6.5 environment to KVM because we can't use > 10gbe speed in our VM's so we decided to test KVM. > > To inicial test I've create script, using qemu commands, to build two VM's > and test the nic speed. > > Each VM has one nic, using virtio, into it own bridge. In this bridge we > associate one 10gbe physical nic. > > They can communicate each other. > > When we try do a copy, through SCP, the speed is 300mbps. And if we open > another copy this speed been the half. > > What we are using: > > S.O: > CentOS 7 > Kernel 3.10.0-123.el7.x86_64 > qemu-kvm.x86_64 10:1.5.3-60.el7 > libvirt-daemon.x86_64 1.1.1-29.el7 > > Hardware: > Dell PowerEdge R720 > 16GB RAM > 2TB SAS > Two Broadcom Corporation NetXtreme II BCM57810 10 Gigabit > > Command to start VM's: > /usr/libexec/qemu-kvm -m 1024 -enable-kvm -k en-us -hda teste1.img -net > nic,model=virtio,macaddr=00:01:02:aa:bb:cc -net > tap,ifname=tap1,script=no,downscript=no -vnc :0 > > What Do I doing wrong? > > Is my test right? > > Is my interpretation right? > > Is my enviroment correct? > > Can I improve or fix any think with my libvirt or command arguments? Have you run the same test on the hosts to see if the problem is the configuration on the host side? In the guests have you looked at mpstat -P ALL while the test was running to see if maybe a core was 100% utilized and you are dealing with a cpu bottleneck? Regards, Dennis _______________________________________________ libvirt-users mailing list libvirt-users@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvirt-users
