Networking & Protocols
Proxy and load balancing: free-recall review
Retrieval beats re-reading. For each prompt, say or write a full answer from memory before you open the model answer — the effort of recall is what makes the material stick.
Reconstruct the unit’s core mechanisms — the two-connection model, the algorithm tradeoffs, active vs passive health, consistent hashing, retry budgets, and the stateless-LB requirement — without looking back at the lessons.
- 01Why does an L7 reverse proxy open two TCP connections per request instead of forwarding packets at layer 3?
- 02Compare round-robin, least-connections, and power-of-two-choices on cost, fairness, and herding.
- 03Why do you need both active and passive health checks, and what does each miss alone?
- 04Why are sticky sessions an anti-pattern, and what is the correct fix? Contrast with consistent hashing.
- 05Explain retry amplification across microservice layers and the mitigations that bound it.
- 06Why must a load balancer behind anycast + BGP ECMP be effectively stateless, and how does Maglev relax that?
If you could reconstruct each answer from memory, you hold the unit’s spine: the two-connection model is what makes HTTP routing and health checks possible; power-of-two-choices is the fair O(1) default that beats both round-robin and least-connections; active plus passive health checks together cover crashes and silent 5xx; sticky sessions are a stateful-backend workaround that Redis replaces (consistent hashing is for cache locality, not sessions); retry budgets, circuit breakers, and jitter bound the cascade at 2× instead of 2^K; and statelessness is what lets anycast + ECMP turn the LB itself from a SPOF into a resilient cluster.