networking question

[Michael Butash]

Agreed, if comcast provides a testing server, test against that first.  Cox
did before, not sure about recently, but speedtest.net isn't bad to use to
test either if you pick a correct server.  They tend to want to test you
against your locale, but as stated, understanding your isp peering egrees
isn't local is important in picking a suitable server.

I was testing out of LA area a customer with Race Communications out of LA
to a good 3-4gb on a local 10gb peer connection there, so I tend to dig for
them when testing locally in phx or ca.  The isp gave me an iperf server to
test to, that I'd hit ~9gbps on a pipe to, so not bad for testing the
interwebs at large.

Otherwise, if your isp is cool with firing up an iperf server too, that's
the real way to test if at line-rate or better of your circuit.  Check tcp
and udp connections, and hold them to their sla's.  Jperf works nice for
gui-based folks, otherwise download iperf3 and rip away when users aren't
on your pipes.  I've hit gig of gig (~993mbps) on windoze using gig usb
nics + iperf even.

-mb

On Wed, Sep 5, 2018 at 9:42 PM, Stephen Partington <cryptworks at gmail.com>
wrote:

> I know for a fact I have a 1gbps connection.
>
> And o e of the issues I ran into is that the ethernet connection between
> the edge device and my ONT would occasionally freak out and drop to
> 100mbps.
>
> I ha e also found that not all edge devices are created equal. I have had
> several devices that at one time were great. We're unable to support full
> gig internet. For testing I have frequently used the Google fiber servers
> to verify speed.
>
>
> On Wed, Sep 5, 2018, 9:35 PM Michael Butash <michael at butash.net> wrote:
>
>> How exactly are you testing your connection?  That's a relevant bit
>> regarding networking.  If testing in phoenix, test a LA California-based
>> server, as most cox residential egresses there.  I like Race Communications
>> out of LA to test against on speedtest.net.  Anything else hits
>> interstate transit, more than cox already does dumping you out LA peering
>> points.
>>
>> Your host nic can handle line rate, and so can the kernel in most cases,
>> really the question is your transport and the applications using it.
>>
>> If using a speedtest.net sort of test, then ymmv with the server you're
>> testing against..  If dealing with comcast, I don't know their peering
>> infrastructure as much as cox's, but they're not typically local, and
>> probably transport you to another state for egress.  This isn't much
>> different from cox, they dump you out LA or Dallas, usually the former than
>> latter, only cbs and other direct peering relations egress actually in
>> phoenix.  Post a traceroute to 8.8.8.8 if you want some recommendations to
>> test against for real world usage.
>>
>> Look at your modem levels too.  You can hit your moto/arris modem on
>> http://192.168.100.1 url to see your modem levels, this has existed
>> since 2000 or so with the advent of docsis and motorola not sucking.  SNR,
>> RX/TX power are what you want to look at, and plenty of wikipedias about
>> docsis to explain tolerances.  If not within tolerance, call the provider
>> to start digging and replace coax as necessary.
>>
>> Every few years my cable gets crappy, and I have to call them to help me
>> dig up and replace bits of either the yard or house to replace.  Arizona is
>> hell on coax with suck-out and other syndromes of extreme heat-based
>> dysfunction.  Expect your levels to go to hell every 3-4 years in arizona
>> with coax.
>>
>> HTH!
>>
>> -mb
>>
>> On Wed, Sep 5, 2018 at 1:12 PM, Jim <jim.nantz15 at comcast.net> wrote:
>>
>>> Back in June comcast raised my connection speed to 150Mbps.  Two weeks
>>> ago it went back down to 100.  I called to complain and was told I was
>>> supposed to be  getting 100Mbps.  I finally got someone to admit that my
>>> connection speed should be 150, but I'm still getting 100.
>>>
>>> I didn't make any changes to the network, but something could have
>>> gotten changed.  Is there anything i can look in network settings to
>>> determine if something could be tweaked for a faster connection?
>>>
>>> I have an Arris SB6141 modem and an ASUS RT-ACRH13 gigabit router.  The
>>> modem (SB6141) supports speeds up to 343Mbps down.  KDE's network manager
>>> shows the MTU setting is set to Automatic and the box is checked to allow
>>> auto negotiation.
>>>
>>> I did install Kubuntu 18 after the increase to 150Mbps.  The speed did
>>> not drop until a month or so after Kubuntu 18 was installed.
>>>
>>> Since the installation, the ethernet connection is no longer called
>>> eth0.  Now it's called enp4s1.  Is that some new way of naming network
>>> devices?  I don't really care what it calls the parts as long as they
>>> work.  I can still use netwatch, but I have to give it the enp4s1 string as
>>> it expects eth0 unless  told differently.
>>>
>>> ---------------------------------------------------
>>> PLUG-discuss mailing list - PLUG-discuss at lists.phxlinux.org
>>> To subscribe, unsubscribe, or to change your mail settings:
>>> http://lists.phxlinux.org/mailman/listinfo/plug-discuss
>>>
>>
>> ---------------------------------------------------
>> PLUG-discuss mailing list - PLUG-discuss at lists.phxlinux.org
>> To subscribe, unsubscribe, or to change your mail settings:
>> http://lists.phxlinux.org/mailman/listinfo/plug-discuss
>
>
> ---------------------------------------------------
> PLUG-discuss mailing list - PLUG-discuss at lists.phxlinux.org
> To subscribe, unsubscribe, or to change your mail settings:
> http://lists.phxlinux.org/mailman/listinfo/plug-discuss
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.phxlinux.org/pipermail/plug-discuss/attachments/20180905/0b34b0cc/attachment.html>