Heavy Networking | May 29, 2015
Show 239 – Design & Build #2 – A New Network From Scratch
Ethan
Banks
Greg
Ferro
Listen, Subscribe & Follow:
Eric Dennington @edennington, Steve Occhiogrosso @StephenO86, and Jason Lavoie @ratvarre join the Packet Pushers to chat about building a new network from scratch. We compiled our notes, and decided on a three phase approach.
Phase 1: Design
- Understand the organization you’re building the network for. You need business requirements.
- Learn about your facilities. New? Old? What’s there already?
- How will legacy be integrated?
- How will the design evolve as needs grow & technology changes in coming years?
- Document the plan. If it’s all in your head, you’re doing it wrong.
- Involve others. Don’t do it alone. Once again, if it’s all in your head, you’re doing it wrong.
Phase 2: Build
- Lab work. Prove your design. Assume nothing.
- Deploy in stages.
- Monitor what you’ve built.
Phase 3: Operate
- Test again after production implementation is final.
- Improve monitoring, especially as unforeseen situations come up.
- Keep communications open with the business and other teams about performance, issues, etc.
Share this episode
Good Podcast guys !!! However, I think there is enough material to talk about building labs specially for guys like me that don’t have the opportunity to see different technologies every day but that needs to keep up with skills and test changes going into production. If you would, consider diving deep into this subject. I am sure that many people are going to thank you for that.
We did touch a bit on having a lab for testing upgrades, config changes and new features. IMO, a good lab setup can be easily re-used and reconfigured to help train on new technologies and do proof-of-concept testing.
I agree, there is plenty of material for a “how to build/run a lab” podcast topic. Shared access is a big one. Maintaining order and structure in an intentionally dynamic environment is another. In my experience, building a lab is a big exercise in compromise, because it is seldom feasible to replicate _all_ real world infrastructure at full scale and one must judge which components can be factored out while still keeping it a valid test.