Up until recently I assumed the azure load balancer just was a simple round robin balancer.
But apparently it is a little more predictive than that. For all practical purposes load balancing can be considered to be random even when connecting from the same client since port reuse on the client is going to mean the hash is different for each request.
But this is still interesting to keep at the back of your head if you ever encounter weirdly balanced load or connection affinity for certain clients since it could be possible to experience that given the load balancer algorithm and certain edge cases.
Post a Comment