[wplug] Verizon vs Comcast
Florin Manolache
florinmano at gmail.com
Tue Feb 11 07:38:58 EST 2014
Rob,
The results are the same while hitting any host from CMU.
Also, http://testmy.net/ (which is not hosted by CMU) shows consistent
results with the CMU tests, as well as netflix and youtube.
So the problem is not on the CMU network for sure.
In fact CMU Computing Services worked to diagnose the problem since it
doesn't affect just the traffic from Verizon, but most of the traffic
through Cogent.
The only thing that came out clearly was the Cogent-Verizon blame game.
In my opinion, even if the real bottleneck is on Cogent's network,
still it is the duty of Verizon to find a way around the problem
because I pay them not Cogent for the bandwidth.
I have reports from Comcast users that they don't see such dramatic
slow downs, but there are no clear measurement results.
That is why I posted the message here, because I can get clear data
from pros, not just tell-tales from regular users.
When you say that you get similar results to me, what exactly are you
referring to?
My results show in the first case a 5 second download, in the second
case a 6 minute download, there is a two orders of magnitude
difference in speed between the two of them, which one did you get?
Thank you for your help,
Florin
On Mon, Feb 10, 2014 at 9:42 PM, Rob <tempest766 at yahoo.com> wrote:
> I get similar results to you and I use neither verizon nor comcast. At this point I'd consider
>
> 1) what kind of delay characteristics are there hitting other machines in the CMU network?
>
> 2) could cogentco be the bottleneck? a bad/weak gateway in their network? lousy load based routing through their network?
>
> 3) is the math.cmu.edu host doing something every evening that slows it down? is the www server configured to throttle based on some activity schedule or system load values? competing with evening network based backups?
>
> Anyway, those are the questions I'd have at this point. I don't think the issue is verizon, comcast, or armstrong (my provider).
>
> -Rob
>
>
>
> --------------------------------------------
> On Mon, 2/10/14, Florin Manolache <florinmano at gmail.com> wrote:
>
> Subject: Re: [wplug] Verizon vs Comcast
> To: "General user list" <wplug at wplug.org>
> Date: Monday, February 10, 2014, 5:20 PM
>
> Rob,
>
> Enclosed is the traceroute listing for normal speed and for
> throttled
> down access, taken
> at less than one hour from each other. I have several sets
> of such
> measurements from different Verizon users.
> Such periods of very low speed appear almost every evening.
>
> What is missing are the results from Comcast users so I can
> better compare.
>
> I'm all for net neutrality.
> My goal here is to vote with my wallet and advice others to
> do the same.
> Even if this will not translate immediately into net
> neutrality, it
> would be a small step in the right direction.
>
> Florin
>
> ----------
>
> Event 1: 8:30pm normal speed
>
> http://my.verizon.com/services/speedtest/
> Download :: 42.1 Mbps Upload :: 31.48 Mbps
> Latency :: 20ms
>
> wget http://www.math.cmu.edu/~florin/SpeedTest/r20m
> 20,000,000 4.53M/s in 4.5s
>
> traceroute to hum (128.2.32.145), 30 hops max, 60 byte
> packets
> 1 192.168.1.1 (192.168.1.1) 1.878 ms
> 1.834 ms 1.872 ms
> 2 L100.PITBPA-VFTTP-03.verizon-gni.net
> (71.245.188.1) 5.890 ms 7.090 ms
> 7.432 ms
> 3 G0-1-1-3.PITBPA-LCR-21.verizon-gni.net
> (130.81.60.172) 12.515 ms
> 12.773 ms 12.779 ms
> 4 xe-6-1-4-0.RES-BB-RTR1.verizon-gni.net
> (130.81.163.72) 80.125 ms
> xe-19-0-8-0.RES-BB-RTR1.verizon-gni.net
> (130.81.209.74) 18.968 ms
> xe-15-1-0-0.RES-BB-RTR1.verizon-gni.net
> (130.81.209.38) 18.149 ms
> 5 0.xe-3-1-0.BR1.IAD8.ALTER.NET (152.63.37.61)
> 22.586 ms
> 0.xe-4-0-0.BR2.IAD8.ALTER.NET (152.63.38.117) 24.630
> ms
> 0.xe-3-1-0.BR1.IAD8.ALTER.NET (152.63.37.61) 23.801
> ms
> 6 te0-6-0-14.ccr41.iad02.atlas.cogentco.com
> (154.54.10.225) 23.207 ms
> 15.376 ms te0-1-0-14.ccr41.iad02.atlas.cogentco.com
> (154.54.12.45) 15.606
> ms
> 7 be2177.ccr22.dca01.atlas.cogentco.com
> (154.54.41.206) 16.838 ms
> be2176.ccr21.dca01.atlas.cogentco.com (154.54.41.54)
> 18.330 ms
> be2112.mpd22.dca01.atlas.cogentco.com (154.54.5.234)
> 19.682 ms
> 8 te4-3.ccr01.pit02.atlas.cogentco.com
> (154.54.30.229) 26.838 ms
> te2-4.ccr01.pit02.atlas.cogentco.com (154.54.3.90)
> 27.422 ms
> te4-3.ccr01.pit02.atlas.cogentco.com (154.54.30.229)
> 28.682 ms
> 9 38.104.121.38 (38.104.121.38) 70.507 ms
> 71.517 ms 72.745 ms
> 10 38.87.51.2 (38.87.51.2) 73.016 ms
> 74.483 ms 74.309 ms
> 11 CORE255-VL987.GW.CMU.NET (128.2.255.249)
> 75.414 ms 76.034 ms 76.332
> ms
> 12 128.2.255.156 (128.2.255.156) 40.404 ms
> 41.586 ms 41.752 ms
> 13 hum.math.cmu.edu (128.2.32.145) 64.275
> ms 64.507 ms 64.468 ms
>
> ---------------
>
> Event 2: 9:15pm very slow network
>
> http://my.verizon.com/services/speedtest/
> Download :: 41.98 Mbps Upload :: 36.97 Mbps
> Latency :: 20ms
>
> wget http://www.math.cmu.edu/~florin/SpeedTest/r20m
> 20,000,000 104K/s in 5m
> 17s
>
> traceroute to hum (128.2.32.145), 30 hops max, 60 byte
> packets
> 1 Wireless_Broadband_Router.home (192.168.1.1)
> 2.018 ms 1.961 ms 2.023
> ms
> 2 L100.PITBPA-VFTTP-03.verizon-gni.net
> (71.245.188.1) 6.383 ms 6.794 ms
> 7.057 ms
> 3 G0-1-1-3.PITBPA-LCR-21.verizon-gni.net
> (130.81.60.172) 10.048 ms
> 10.376 ms 10.365 ms
> 4 xe-6-1-4-0.RES-BB-RTR1.verizon-gni.net
> (130.81.163.72) 14.790 ms
> xe-21-1-2-0.RES-BB-RTR1.verizon-gni.net
> (130.81.163.68) 16.251 ms 17.360
> ms
> 5 0.xe-2-3-0.BR1.IAD8.ALTER.NET (152.63.7.229)
> 17.951 ms 16.505 ms
> 16.493 ms
> 6 te0-3-0-14.ccr41.iad02.atlas.cogentco.com
> (154.54.13.137) 57.990 ms
> 54.863 ms *
> 7 be2112.mpd22.dca01.atlas.cogentco.com
> (154.54.5.234) 58.009 ms 57.926
> ms 56.749 ms
> 8 te2-4.ccr01.pit02.atlas.cogentco.com
> (154.54.3.90) 68.463 ms
> te4-3.ccr01.pit02.atlas.cogentco.com (154.54.30.229)
> 68.388 ms
> te2-4.ccr01.pit02.atlas.cogentco.com (154.54.3.90)
> 67.513 ms
> 9 38.104.121.38 (38.104.121.38) 69.373 ms
> 69.624 ms 69.571 ms
> 10 38.87.51.2 (38.87.51.2) 76.202 ms
> 76.296 ms 76.281 ms
> 11 * CORE255-VL987.GW.CMU.NET (128.2.255.249)
> 68.027 ms 68.179 ms
> 12 128.2.255.156 (128.2.255.156) 57.142 ms
> 57.525 ms 58.291 ms
> 13 hum.math.cmu.edu (128.2.32.145) 65.374
> ms 65.381 ms 65.849 ms
>
> On Mon, Feb 10, 2014 at 12:50 PM, Rob <tempest766 at yahoo.com>
> wrote:
> > I'd be very careful about using anecdotal evidence to
> say that one evil carrier (verizon or comcast) is better
> than the other. The truth is that without better
> testing you don't kow who or where the problem is
> occuring. It could be that the bastards are taking
> advantage of their privileged status of information and data
> provider to play favorites in routing or prioritizing
> traffic. The solution there is to support net
> neutrallity andtell the FCC that you want all ISPs and
> network providers to be designated common carriers and as
> public utilities so that they cannot play favorites.
> >
> > I consider both verizon and comcast to be evil but
> alas, they have a relatively captive audience with little to
> no real competition or customer choice.
> >
> > post your traceroute data taken at several times during
> a 72 hour period and we'll take a look to see if anything
> jumps out as a bottleneck.
> >
> > -Rob
> >
> > --------------------------------------------
> > On Mon, 2/10/14, Florin Manolache <florinmano at gmail.com>
> wrote:
> >
> > Subject: [wplug] Verizon vs Comcast
> > To: "General user list" <wplug at wplug.org>
> > Date: Monday, February 10, 2014, 10:04 AM
> >
> > Hi everybody,
> >
> > I know this is not a Linux specific question, but
> I think
> > this is a
> > good place to dip into some expertise.
> >
> > The problem: I am working at CMU and use Verizon
> FiOS at
> > home (50Mbps download).
> > During the late evenings the download speed to
> CMU goes
> > under 1Mbps,
> > sometimes much lower.
> > Bandwidth tests show that locally in the Verizon
> network the
> > access
> > speed is fine (around 35-40Mbps) but once the
> traffic gets
> > out of
> > Verizon's network it is throttled down.
> > Verizon blames Cogent for that, Cogent blames
> Verizon, and
> > the end
> > user is the one taking the hit.
> > This is not only my problem, but all my
> colleagues on
> > Verizon FiOS
> > experience it.
> > Also, not only the traffic from CMU gets
> throttled down, but
> > a lot of
> > other sites, Netflix, YouTube, etc, basically
> everything
> > that follows
> > the Verizon-Cogent route.
> >
> > Potential solution: some people claim that the
> problem
> > doesn't exist
> > from the Comcast network.
> > Even if the bandwidth is lower than the one
> offered by
> > Verizon, the
> > claim is that Comacat's bandwidth is sustainable
> at the
> > order of about
> > 20Mbps even during the evenings.
> > If I could check that, I'll switch to Comcast and
> advice all
> > my
> > Verizon FiOS colleagues to do so.
> >
> > I would like to hear your opinions, thoughts,
> and
> > experiences
> > regarding this issue.
> > If you have Comcast please try some bandwidth
> tests around
> > 9-10pm and
> > let me know the results.
> > For the bandwidth tests I was using http://testmy.net/ or a 20MB
> > download from one of my servers at
> > http://www.math.cmu.edu/~florin/SpeedTest/r20m
> >
> >
> > Thank you very much,
> > Florin Manolache
> > _______________________________________________
> > wplug mailing list
> > wplug at wplug.org
> > http://www.wplug.org/mailman/listinfo/wplug
> >
> > _______________________________________________
> > wplug mailing list
> > wplug at wplug.org
> > http://www.wplug.org/mailman/listinfo/wplug
> _______________________________________________
> wplug mailing list
> wplug at wplug.org
> http://www.wplug.org/mailman/listinfo/wplug
>
> _______________________________________________
> wplug mailing list
> wplug at wplug.org
> http://www.wplug.org/mailman/listinfo/wplug
More information about the wplug
mailing list