We are current using 2 vlm-1000s and 2 virtual loadmaster geo systems in our setup. There are two subnets (2 data centers connected with fiber but 2 miles apart and <1ms latency) and in each dc is a sharepoint stack. So each side has 1 load balancer, 1 geo, 2 web front ends.
We have round robin on everything so that the geos are round robin to the load balancers which in turn are round robin to the web front-ends. In this scenario either subnet will respond. Our entire farm is virtual including the dbs and since we are using always on this allows us to be fully HA.
Its is working great so far and I am wondering if anyone is doing anything similar or if anyone can point and flaws in this configuration scenario.