Deployment & Infra
Load balancing levels: 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 layer model and its operational knobs stick.
Reconstruct the unit’s spine — what each layer can see, why TLS termination is the cleanest lens on the split, and how algorithms, health checks, affinity, and draining decide reliability — without looking back at the lesson.
- 01Why can an L4 balancer route on IP and port but never on URL path or HTTP header, and what changes at L7?
- 02Explain why TLS termination is the cleanest lens on the L4/L7 split, including what an L4 balancer does with TLS instead.
- 03Round-robin vs least-connections vs hashing — when does each win, and what does hashing give an L4 balancer?
- 04What do active vs passive health checks do, and why does the check interval bound how fast a dead backend is removed?
- 05What is connection draining, what breaks without it, and what is the AWS default?
- 06Why do sticky sessions hurt even when they work, and what is the senior reflex?
If you could reconstruct each answer from memory, you hold the unit’s spine: the layer decides what the balancer can see (L4 = IP/port and bytes, L7 = parsed HTTP), TLS termination is the cleanest lens on that split (decrypt to route on path, or pass through to stay blind and end-to-end), algorithms match the request profile (least-connections under uneven durations, hashing for L4 affinity), health checks pull dead backends only as fast as their interval, connection draining (AWS default 300 s) lets in-flight requests finish before a backend retires, and sticky sessions trade even distribution and clean draining for pinned state — so externalise state first.