Monitoring your gateway is fine. If it's something that is generally very steady over long periods of time, like as shown in the quality graph, and only changes when you change things local to your end, then it is probably safe to pinpoint that back to local changes you've made. I didn't realize you were referring to the quality graph over long periods of time, sounded like you pinged things on occasion and were accounting a 1-2 ms change as something you did - even for your gateway you'll commonly see more than 1-2 ms variance from one time of day to another depending on many different factors, but that won't necessarily always be the case. Checking a ping time on occasion is much different than comparing repeated ping history like the RRD graph shows. So you probably do have that kind of difference from going to ESX in that case. Why I don't know, there isn't that much difference generally. Pinging from the physical server this site runs on, through a firewall in ESX, out of ESX up to the datacenter's router, adds 0.2-0.3 ms vs. pinging the LAN IP of the firewall (and has response time in the neighborhood of 0.5 ms, close to what LAN to LAN pings commonly are), and that's nothing more than adding the ~0.2-0.3 ms response time from the firewall's WAN to that router. That's more or less the same as a fully physical network would see, so it's not typical of ESX.
