WeightedRoundRobin LeastConnections
Refinements on plain round robin for load balancing: weighted round robin sends more traffic to higher-capacity servers, least-connections sends the next request to whichever server currently has the fewest in-flight requests.
Why we need this / what value this brings
Fixes plain round robin's blind spot: it sends equal traffic to unequal-capacity servers, or to servers already busy with slow requests.
When to use this
Once backend instances have different capacity, or request processing time varies enough that 'currently busiest server' matters more than 'whose turn is it'.
How to use or implement this
Least-connections is usually the safer default once instances vary at all — it adapts to actual load rather than requiring you to keep capacity weights manually accurate.
Research questions
- Only relevant once backend instances differ in capacity or request cost varies a lot — plain round robin is fine for uniform instances and uniform request cost.
Empty folder — drop notes, links, and findings here as you research.