Descriptions of the Weighted Balance algorithm here and in the knowledgebase refer to “amounts” and “proportions” of traffic, but I haven’t found a clear description of what is measured.
I’ve found that if I weight two comparable links equally, the first one will end up wth almost twice as many bytes of data as the second. I suspect because the algorithm is counting or alternating sessions or something similar, rather than counting bytes, and initial sessions are likely often larger than subsequent sessions. As a a result, I’ve had to adjust the weighted balances to somewhere between 5:10 and 7:10 to account for that. Any clarity on how traffic is actually measured for this algorithm?