Google SRE Prodcast Maglev: load balancing at Google with Cody Smith and Trisha Weir
12 snips
Nov 13, 2024 Cody Smith, CTO and co-founder of Camu Energy, spent over 14 years at Google and contributed to Maglev. Trisha Weir, with 21 years at Google, is an SRE Department Lead. They uncover the evolution of Maglev, a network load balancer essential for traffic management in data centers. Their discussion highlights the significance of psychological safety and collaboration in tech innovation. They also delve into challenges faced during system rollouts, debugging practices, and the shift from manual to automated network provisioning, showcasing a unique blend of technical and teamwork insights.
AI Snips
Chapters
Transcript
Episode notes
From Student ISP To SRE Careers
- Cody and Trisha met running a student ISP at UC Berkeley and that early hands-on work shaped their SRE careers.
- Trisha described wiring buildings and caring deeply about users when connections failed.
Load Balancing Is Front End Complexity
- Front-end infrastructure is the complex shared stack from user request to backend handling.
- It requires DNS, global load balancing, failover, and DDoS absorption to maintain reliability.
Vendor Hardware Became A Bottleneck
- Google used expensive vendor network load balancers that cost roughly $20,000 per box with pricey support.
- The devices became a bottleneck and didn't match Google's scaling and feature needs.