I obscured Comcast hostnames that more specifically pinpoint my location in FL, other than that these are not edited in any way. Since I just started recording data a few hours ago, this is not a complete picture, but it is a picture of what a few hours of sampling every minute with traceroutes is demonstrating about AT&T's network leading in to Blizzard's.
* Using xmlroute to run traceroutes to 22.214.171.124. This is not the "official" IP I saw in some other threads, but it is the IP that the Diablo III process connects to consistently when viewing Windows' TCP connections.
* Using background Windows task to run xmlroute and pipe output to a file on disk every minute.
* Using Splunk to pickup the data in that file, with each hop delimited as an event.
* Using Splunk queries to extract the data and visualize it as seen in the imgur url.
In general, today is a reasonably good latency day for me, though it IS a bit spikey. I can fire nether tentacles right now and won't see it render for a second or two every few minutes. The important take away there is that where you see high latency in my graphs, you should know that the actual latency is quite a bit higher thanks to averaging.
In the horizontal bar chart, the bottom 4 hosts are my router and Comcast's routers. You can see that over the last few hours, not only is ATT spikey, but it's also 2-3x worse than Comcast.
Problematic hosts in AT&T's network:
When you look at my timechart you can see some hosts do spike on occasion, but gar29.la2ca.ip.att.net is a consistently bad performer over time.
Just wanted to bring in some data over time into the latency discussion as the spikey nature of ATT's latency sometimes prevents point-in-time traceroutes from revealing a complete picture.
For the blues, should we as end-users be reaching out to ATT on this issue? I assume Blizzard knows where packet loss or congestion is happening and is working with ATT on it.