NFVI Upgrades and NFV Migrations with Live Critical Telco
summary
Two years of lab work produce a repeatable NFVI CI/CD program that performs NFVI upgrades with live traffic (OpenStack VMs and CNFs on Kubernetes) every ~3 months, runs resilience tests (reboots, switch failures, migrations), and gates releases to GA. Key findings: whole-datacenter upgrades can take months, backups often restore slower than redeploy, firmware/driver order matters, and upgrades expose preexisting resilience holes. Operational fixes include reboot alignment and strict pod disruption budget planning.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 0:22Key takeaways
- 2:27Who is Gerardo Martinez
- 5:11Why do we upgrade
- 8:40The Ericsson solution
- 10:24Telco facts and practices
- 15:56Summary
- 17:08Challenges
- 21:32Migrations
- 27:56Pet vs Cattle
related talks
24mCI-CT-CD, Devops loop for future Telco Cloud
50mOps issues
39mVirtualization, Automation and DevOps – Why it’s Taking so long for Telcos?
14mDeep Optimization of VMM Live Upgrade
21mHardware Automation Deployment and Test with Big Scale in ChinaMobile NFV cloud
37mZero-Regression Network CI/CD for Finance-Grade Reliability
15mKeeping a Hypervisor Fleet Up to Date with Minimal Customer Disruption
39mLessons Learned From Maintaining Continuous Delivery While Migrating From a Static Infrastructure to Kubernetes [B]