@devopsec: at the end you get more or less to a specific order of selecting the addresses to route to. For example, with two addresses one 99 and the other 1, the one with weight 1 is going to be placed the 2nd (or 1st depending on the initial order). If weights are 90 and 10, it should end up like 1st, 2nd, 1st, 1st, 1st, 1st, 1st, 1st, 1st, 1st, 1st, 2nd, 1st, 1st.... For calls of same type (e.g., codecs, audio/video) and duration, could bring some benefits, but for calls of random type and duration the random ordering proved to give pretty good load of resources.

With the above remarks, considering that random ordering was done on purpose at the time of implementation, I don't consider this PR a bug fix, but a different distribution algorithm (for example it can be called percentage ordered distribution).

—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.