Keeping the Balance: Load Balancing Demystified
summary
Explains concrete choices and tradeoffs: an example server handles five queries per second, DNS round‑robin adds a TTL delay (short TTLs multiply DNS load ~12x), network load balancing uses five-tuple hashing and causes hotspotting from elephant flows, direct server return (DSR/GRE encapsulation) preserves client IP and shifts outbound bandwidth off the VIP, anycast routes risk cascading failures, and service‑mesh sidecars replace fat clients to centralize load signals and retries — each approach works for specific latency, control, and scaling constraints.
AI-written summary · about
chapters (25)
- 0:00Intro
- 1:31Why is load balancing important
- 2:09Example
- 7:38DNS TTL
- 9:13DNS Load Bouncing
- 10:12Load Balancing
- 11:35Five Tuple hashing
- 12:48Lair for Balancer
- 14:01Load Distribution
- 15:20Network Load Balancing
- 22:33Web Services
- 23:57Layer 2 DSR
- 24:59Layer 3 DSR
- 27:23Multiregion
- 31:04Cover Routes
- 33:40DNS Geo Load Balancing
- 35:06DNS Extensions
- 35:52Back to our story
- 36:21Layer 7 Load Balancing
- 39:25Algorithms
- 40:46Client Load Balancing
- 42:52Service Mesh
- 44:21Consistency
- 44:36Takeaways
- 45:02On Time
related talks
30mKeeping the Balance: Internet-Scale Loadbalancing Demystified
29mLoad balancing: an in-depth study to scale @ 80K TPS
53mHTTP Traffic Load Balancing
29mUsing the Internet as Your Load-Balancer
41mLoad Balancing Roundup
40mGood, Bad, and Ugly - the Art of Load Balancing at Scale
52mGlobal web resource load balancing across geo‑distributed infrastructure
13mEdgeBalance: Model-Based Load Balancing for Network Edge Data Planes