Skip to content

Instantly share code, notes, and snippets.

@chenchun

chenchun/Maglev.md

Last active November 18, 2019 02:38
Show Gist options
  • Save chenchun/d04bc7123a84bf865404597e8a7e40c6 to your computer and use it in GitHub Desktop.
Save chenchun/d04bc7123a84bf865404597e8a7e40c6 to your computer and use it in GitHub Desktop.
Download ZIP
load balance #loadbalance #lb
Raw
Maglev.md

https://ai.google/research/pubs/pub44824 https://simpleyyt.com/2017/07/05/introduction-to-google-maglev/

  • router ECMP
  • LB-RS GRE
  • Direct Server Return(DSR), send VIP->client IP to router, bypass LB
  • bypass kernel, DPDK
  • session affinity, consistent hashing

Ucloud vortex http://www.infoq.com/cn/articles/Maglev-Vortex

  • DR pure L2 network

meituan https://tech.meituan.com/MGW.html

  • global session table

performance

  • linux内核lvs处理包的速度太慢（10G网卡3～4Mpps），特别是对小包的处理，导致了性能瓶颈
  • 利用DPDK技术可以达到线速（10G网卡14Mpps）
Raw
opensource.md
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment