To date, Trellis is the only production-ready, open source, whitebox-based SDN leaf-spine fabric solution designed for telco service provider’s edge cloud. In this talk, we are going to share technical issues and challenges we have been facing in large scale production deployment, the techniques we used to debug those issues and provide a few suggestions for developers who want to build scalable SDN applications.

12 lessons we learnt from production deployment:

  • Use distributed primitives & threads wisely
  • Use iterations wisely
  • Know your environment well
  • Know your programming language well
  • Profiler is your friend, for most of the time
  • Log is not always your friend
  • Question everything
  • Protect the system from cascading failures
  • Never judge based on a single input
  • Watch for early warnings
  • Always have a Plan B
  • Know how it went south

Date: Apr 22, 2019
Event: SDNDS-TW/CNTUG Meetup #15

Categories: Talks