[dpdk-dev] More performance comparisons on virtual switches
jun.xiao at cloudnetengine.com
Sat Aug 15 18:11:59 CEST 2015
Since we announced CloudNetEngine virtual switch technical preview, we received quite a few very good suggestions/feedbacks from evaluation customers, thank you all!
Today we release CloudNetEngine vSwitch technical preview update 1, which includes a number of performance improvements over last few weeks. Here we also share more performance comparisons on Native kernel OVS, OVS-DPDK, and CNE vSwitch technical preview update 1. The methodology for the comparison is also commonly used by virtualization vendors, i.e. end to end TCP throughput/CPU usage, and TCP_RR transaction throughput.
A few key observations:
- OVS-DPDK is not designed to optimize for intra VMs communication, its tcp throughput is even worse than Native kernel OVS.
- OVS-DPDK only wins at TCP_RR, and this is reasonable as it always keep a CPU core 100% busy.
- CNE vSwitch wins on almost all metrics (except OVS_DPDK's TCP_RR), and its TCP throughput/CPU efficiency is around 3X against Native kernel OVS while also provides better TCP_RR throughput than Native kernel OVS.
For more information, please refer to http://cloudnetengine.com/en/blog/2015/08/15/more/
If you want to try CNE vSwitch technical preview update 1, please don't hesitate to send a evaluation request to info at cloudnetengine.com.
More information about the dev