Split traffic across weighted origin pools and see the blended latency, plus what happens when a pool drops out.
Weighted routing normalises against the sum of healthy weights, so ejecting a pool silently redistributes its share to the survivors in proportion to their own weights rather than evenly. Blended latency is the weighted mean, which is why sending even 10% of traffic to a distant pool moves the average far less than it moves that 10% of users. The blended number is what your dashboard shows and the per-pool number is what those users feel, so weighted routing can look healthy while a tenth of your traffic is served twice as slowly — and the failure case shows what happens when the nearest pool is the one that drops.
Traffic Routing
pool share = pool weight ÷ sum of healthy weights; blended latency = Σ(weight × pool latency) ÷ sum of healthy weights, with unhealthy weights removed before normalising.
pool share = pool weight ÷ sum of healthy weights; blended latency = Σ(weight × pool latency) ÷ sum of healthy weights, with unhealthy weights removed before normalising. Weighted routing normalises against the sum of healthy weights, so ejecting a pool silently redistributes its share to the survivors in proportion to their own weights rather than evenly. Blended latency is the weighted mean, which is why sending even 10% of traffic to a distant pool moves the average far less than it moves that 10% of users.
The blended number is what your dashboard shows and the per-pool number is what those users feel, so weighted routing can look healthy while a tenth of your traffic is served twice as slowly — and the failure case shows what happens when the nearest pool is the one that drops.
This calculator takes 8 inputs: Requests per month, Pool A weight, Pool B weight, Pool C weight, Pool A latency, Pool B latency, Pool C latency, Pool health. The pre-filled defaults are a realistic starting point — replace them with figures from your own environment for a result you can act on.