CloudNetEngine vSwitch 2.1 boosts virtual network performance further


CloudNetEngine vSwitch 2.1 uses a very unique network stack to serve both NFV workloads and Enterprise workloads with the most high performance in the industry.

Posted by CloudNetEngine on April 25, 2018

What's new in CloudNetEngine vSwitch 2.1

  • Pure polling RX mode to improve NFV workloads performance further. (CNE vSwitch uses interrupt + poll hybrid RX mode by default)
  • A new scheduling policy for RXQ NUMA strong affinity.
  • Performance improvement on large scale concurrent flows.
  • Upgrade control/management front end to OVS 2.8.2.

Features in CloudNetEngine vSwitch 2.0

  • High reliability dataplane
  • RXQ dynamically load balancing among available worker threads
  • Hybrid RX mode with interrupt + poll
  • VXLAN, GENEVE, GRE tunneling
  • VLAN/QinQ/TSO/UFO/CKSUM SW processing or NIC acceleration
  • LRO
  • TX zero copy
  • Multiqueue
  • Conntrack
  • NAT
  • L3 reassembly
  • Physical NIC bonding
  • Jumbo Frame
  • Tcpdump

vSwitch performance comparison

2.1-1.png

For NFV workloads, CNE vSwitch 2.1 can get 20~50% higher throughput than OVS-DPDK under various packet size and number of concurrent flows.

2.1-2.png

 

For Enterprise workload, CNE vSwitch 2.1 saves around 50% CPU cycles and get 10~20% higher throughput comparing to Native OVS.

2.1-3.png

2.1-4.png

2.1-5.png

 

For more information, please contact info@cloudnetengine.com.