Analyse performance gaps with QEMU's virtio-net implementation #369
Description
Non-vhost, not-multiqeue QEMU:
qemu-system-x86_64 -machine q35,accel=kvm,kernel_irqchip -cpu host -m 512 -bios ~/Downloads/OVMF.fd -device virtio-blk-pci,drive=root -drive if=none,id=root,file=/home/rob/workloads/clear-30800-kvm.img -nodefaults -device virtio-net-pci,netdev=mynet0 -netdev tap,id=mynet0,ifname=tap0,script=no,downscript=no,vhost=off -serial stdio
iperf3 result:
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-10.00 sec 26.3 GBytes 22.6 Gbits/sec 0 sender
[ 5] 0.00-10.00 sec 26.3 GBytes 22.6 Gbits/sec receiver
vs CH with virtio-net:
target/debug/cloud-hypervisor --kernel ~/src/rust-hypervisor-firmware/target/target/release/hypervisor-fw --serial tty --console off --disk path=~/workloads/clear-30800-kvm.img --net tap=tap0,mac=12:34:56:78:90:ab --memory size=512M
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-10.00 sec 9.25 GBytes 7.94 Gbits/sec 0 sender
[ 5] 0.00-10.00 sec 9.25 GBytes 7.94 Gbits/sec receiver
i.e. Cloud Hypervisor's peformance is ~ 35% of QEMU.
Metadata
Metadata
Assignees
Labels
No labels