Loading…
September 11-14, 2017 - Los Angeles, CA
Click Here For Information & Registration
View analytic
Thursday, September 14 • 3:30pm - 4:10pm
Improve the performance of GTP-U and Kube-Proxy Using VPP - Hongjun Ni, Intel

Sign up or log in to save this to your schedule and see who's attending!

Come to this combined session where you will hear about:

Improve the Performance of Kubernetes Cluster with VPP Based Router and Load Balancer
In current kubernetes environment, the container network’s data paths are all in kernel space:
Container connects to the host with Linux virtual Ethernet device (VETH), kernel routing rules
And iptables rules enables the connection of the container from/to other containers or external Hosts; the default (which gains best performance) load balancer for the kube-proxy are also Implemented with iptables. This presentation will give another approach to enable the container Network with an userspace application named VPP to archive much better performance, two key technologies are used:
1> Virtio_user based data path to archive best performance for virtual NICs.
2> VPP load balancer to enable load balancer in userspace to archive better performance for kube-proxy

Six Ways to Improve Data Plane Performance in VPP and Container
VPP is a high performance, packet-processing stack that can run on commodity CPUs. But for some developers, if they did not know some tricks, they could not develop high performance data plane using VPP. The presentation will show six ways to optimize the data plane performance in VPP and container.It includes below key elements:

1) One-loop, Dual-loop and Quad-loop and performance evaluation.
2) Packet prefetching and performance evaluation.
3) Bypass first routing lookup and performance evaluation.
4) Bypass second routing lookup and performance evaluation.
5) Handle packet header in advance and performance evaluation.
6) Tweak startup configuration in container and performance evaluation.

Speakers
avatar for Hongjun Ni

Hongjun Ni

Senior Software Engineer, Intel Corporation
Hongjun Ni is a Senior Software Engineer at Intel, focusing on high performance data plane and SDN/NFV. | He is FD.io subproject VPP Maintainer, Hc2vpp committer and NSH_SFC committer. | He has twelve years' production experience on developing SDN, NodeB, RNC, SGSN, GGSN, FMC(Fix... Read More →


Thursday September 14, 2017 3:30pm - 4:10pm
Gold 3
Feedback form isn't open yet.

Attendees (5)