this isn't exactly a linux question, although it is related to this tc stuff i've been messing around with lately...
this java-based test
http://netalyzr.icsi.berkeley.edu/
is one i found recently that purports to test numerous aspects of one's network.
normally, when I would run it, the "network buffering" measurements were good; on my 50 down/10 up Comcast business connection, through my linux iptables router PC, the "downlink" would have no recorded latency, and the "uplink" would typically be around 220ms, which they say is good.
now, recently i have used tc to limit the download/upload speed on the external interface of this router. I set the download to a max of ~52mbps, and the upload to a max of ~10.5, as I noticed that Comcast tends to give a few megabits over what they say they will in this after.
upon using this test again, my buffer measurements were 170ms on the downlink and a large 970ms on the uplink... ?
Idk if this test is full of crap, but a nearly 1-second jump in latency seems quite high, particularly as limiting your router's speeds to slightly under what your ISP gives you appears, through googling, to be a common resolution for buffer-related issues.