dslreports logo
 
    All Forums Hot Topics Gallery
spc
Search similar:


uniqs
979

ropeguru
Premium Member
join:2001-01-25
Mechanicsville, VA

ropeguru

Premium Member

Question about Comcast's speedtest sites

So the other day I was using Comcast's speed test sites just to see what kind of rates I get from different parts of the Country.

One thing I found unusual was that the latency showing during the test was really high in some cases. For example when the test started from the Miami site, the test was showing 355ms of latency. So me being the curious type, I fired up wireshark and determined the IP that the test was running from so I could do a trace route to see where the latency issue was occurring. To my surprise, I had less than 50ms latency to the IP the test was being run from.

So my questions i just how is the latency for the speed test determined? Anyone know?

Extide
join:2000-06-11
Salt Lake City, UT

Extide

Member

Those flash based speed tests are notoriously in-accurate, so it is probably just a badly measured latency number.

bradyr
YCCD - Network Operations
Premium Member
join:2008-10-27
Sonora, CA
(Software) pfSense
Netgear CM1200
Ubiquiti UAP-AC-HD

bradyr to ropeguru

Premium Member

to ropeguru
if you want to do a real speed test, find a NDT server in your area and test against that (lots of universities host them). if you're looking for a good speed test that takes your powerboost into account, do the shaperprobe thing.

otherwise, just download a big file from somewhere like microsoft or grab one of the linux distro's. use a calculator and determine your own real-world speeds.

I also wouldn't worry about the latency reports on the speedtest servers, either. if you're concerned about latency, pick some of your favorite websites and do some ping and trace-route testing.

whfsdude
Premium Member
join:2003-04-05
Washington, DC

whfsdude to ropeguru

Premium Member

to ropeguru
NDT or Iperf

andyross
MVM
join:2003-05-04
Aurora, IL

andyross to ropeguru

MVM

to ropeguru
Does it always do it? It could have been a brief busy time when it was checking the latency.

ropeguru
Premium Member
join:2001-01-25
Mechanicsville, VA

ropeguru

Premium Member

said by andyross:

Does it always do it? It could have been a brief busy time when it was checking the latency.

At the time I was testing I got the same results over a 10 minute period. And during each test I had a ping going to the test server and when the test would run, the ping would be in the 50ms range but the test was saying 300+ms.

NetFixer
From My Cold Dead Hands
Premium Member
join:2004-06-24
The Boro
Netgear CM500
Pace 5268AC
TRENDnet TEW-829DRU

NetFixer

Premium Member

said by ropeguru:

said by andyross:

Does it always do it? It could have been a brief busy time when it was checking the latency.

At the time I was testing I got the same results over a 10 minute period. And during each test I had a ping going to the test server and when the test would run, the ping would be in the 50ms range but the test was saying 300+ms.

The difference is likely that the test is using imbedded results taken from inside the HTTP/TCP packets used in the test itself, and the latency reported by the test would depend greatly on how much load that server was under (you probably aren't the only one running a test at any given time) Your ping test is using the ICMP echo which requires very little of the server's resources.

YukonHawk
join:2001-01-07
Patterson, NY

YukonHawk to ropeguru

Member

to ropeguru
Here an NDT server site run bu Rutgers

»ndt-nbp.rutgers.edu/