Fiber optic speed test.

I just got FIOS yesterday, so today I ran speed tests.

Verizons says I have 3/1 (no decimal places iirc) and that's what I paid
for. But after I ran it a few times with Ookla, it said it was 4.1 down
/ 1.9 up. My friend told me that, like a car on a cold morning, it's
because after you run it a few times you get the wires warmed up and
they work better. How many times should I run it to get the best
possible test?

On 2019-11-13 11:53 a.m., micky wrote:
I just got FIOS yesterday, so today I ran speed tests.

Verizons says I have 3/1 (no decimal places iirc) and that's what I paid
for. But after I ran it a few times with Ookla, it said it was 4.1 down
/ 1.9 up. My friend told me that, like a car on a cold morning, it's
because after you run it a few times you get the wires warmed up and
they work better. How many times should I run it to get the best
possible test?


You do not have to Warm it up! that is FUD

your numbers make no sense, Try another test .
My system on cable gives me Ping 9, 331.72 Mb/s down and 16.53 Mb/s up
using OOkla, you should have a similar format but different numbers.

Rene

In article , micky
wrote:

I just got FIOS yesterday, so today I ran speed tests.

Verizons says I have 3/1 (no decimal places iirc) and that's what I paid
for. But after I ran it a few times with Ookla, it said it was 4.1 down
/ 1.9 up.

that ain't fios.

fios currently has a minimum speed of 100 mbit symmetrical for new
subscribers.

3/1 sounds like dsl.

My friend told me that, like a car on a cold morning, it's
because after you run it a few times you get the wires warmed up and
they work better.

he's either playing you or he's clueless.

How many times should I run it to get the best
possible test?

once is sufficient.

maybe a second time later in the day to make sure the numbers are
consistent.

On 11/13/2019 10:53 AM, micky wrote:
I just got FIOS yesterday, so today I ran speed tests.

Verizons says I have 3/1 (no decimal places iirc) and that's what I paid
for. But after I ran it a few times with Ookla, it said it was 4.1 down
/ 1.9 up. My friend told me that, like a car on a cold morning, it's
because after you run it a few times you get the wires warmed up and
they work better. How many times should I run it to get the best
possible test?



Your friend is completely wrong. You can run it multiple times and get
different results each time--some faster and some slower. That's because
conditions keep changing.

Are those numbers Gigabytes per second? I'm envious. I wish I could get
FIOS, but it's not available here.

--
Ken

In article , Ken Blake
wrote:

On 11/13/2019 10:53 AM, micky wrote:
I just got FIOS yesterday, so today I ran speed tests.

Verizons says I have 3/1 (no decimal places iirc) and that's what I paid
for. But after I ran it a few times with Ookla, it said it was 4.1 down
/ 1.9 up. My friend told me that, like a car on a cold morning, it's
because after you run it a few times you get the wires warmed up and
they work better. How many times should I run it to get the best
possible test?


Your friend is completely wrong. You can run it multiple times and get
different results each time--some faster and some slower. That's because
conditions keep changing.

they won't be exactly the same, however, they will be fairly close,
other than if there's a major problem or unusually high demand.

Are those numbers Gigabytes per second? I'm envious. I wish I could get
FIOS, but it's not available here.

they aren't gigabit.

it's mbit, and dsl speeds.

currently, fios is 100 mbit minimum, up to 1 gigabit.

On Wed, 13 Nov 2019 12:53:36 -0500, micky
wrote:

I just got FIOS yesterday, so today I ran speed tests.

Verizons says I have 3/1 (no decimal places iirc) and that's what I paid
for. But after I ran it a few times with Ookla, it said it was 4.1 down
/ 1.9 up. My friend told me that, like a car on a cold morning, it's
because after you run it a few times you get the wires warmed up and
they work better. How many times should I run it to get the best
possible test?

What nospam said. That's not FIOS and someone is playing you regarding the
wires warming up.

I would only add that each of your speed tests is a snapshot in time. The
first test of the day might be the fastest, or the 2nd or 3rd or whatever.
If you're curious, run a speed test at multiple times of the day and on
multiple days. Over time, you'll see what is normal for your situation.

micky wrote:

I just got FIOS yesterday, so today I ran speed tests.

Verizons says I have 3/1 (no decimal places iirc) and that's what I
paid for. But after I ran it a few times with Ookla, it said it was
4.1 down / 1.9 up.

4.1 and 1.9 are just numerical values with no magnitude, like Kbps,
Mbps, or Gbps. That's like saying you car is moving at 50. 50 what?
Could be miles per hour, kilometers per hour, knots (if you have one of
those amphibious cars), or some other magnitude. 3/1 could be 3 Kbps up
and 1 Kbps down, but that's worse than dial-up service. Could be 3 Mbps
down and 1 Mbps up, but that would be damn slow even for copper
connects. While fiber only runs up to the service entry to your home or
maybe just to the doghouse station in the neighborhood, the rest is
still copper. I would think FIOS should give 3 Gbps down and 1 Gbps up.
What /speeds/ is the 3/1 plan supposed to give you?

I typically run a speed test about 5 times. That's to try to get rid of
any caching effects on the first one or two runs and account for any
variation in how busy are the nodes (hops) in the route between me and
the test server.

Using your own ISP's speed test server is more accurate then using
something outside their network (e.g., speedtest.net by Ookla). Your
ISP only has responsibility for trying to provide you with the service
tier for which you pay them, not for anything outside their network.

My friend told me that, like a car on a cold morning, it's because
after you run it a few times you get the wires warmed up and they
work better. How many times should I run it to get the best
possible test?

"wires warmed up". Did you actually believe that? Your friend must
think you're stupid, or he's attempting a pretense of expertise. I
remember fixing my dad's home PC (used for his SOHO business) and was
making some fixes when he asked me why I did something. I replied,
"Because it's Tuesday". He looked puzzled at first (because I wasn't
grinning nor even looked at him), but then figured out I was telling him
to shut up in a roundabout way. He wouldn't understand if I explained
it to him, anyway.

You don't have synchronous service (downstream and upstream bandwidth
are the same). Instead you have asynchronous service (downstream and
upstream bandwidth are not equal). They give higher downstream
bandwidth because the vast majority of consumer traffic is to download
content from web sites, not to operate servers to which outsider can
connect (which may violate their terms of service for a non-business
service tier). Users far less often upload files or run servers for
outsiders to connect to, so upstream bandwidth is far less.

Also, just because you have a larger pipe to your home doesn't mean
servers are going to speed up. They have a limited number of resources,
like concurrent connections, that have to get shared with all visitors.
You cannot draw content faster from them then they will deliver despite
how fat is the network pipe to your home. Think of like your sink. The
amount of water that comes out is still controlled by the faucet. You
might expect a huge speed increase but not actually experience nearly as
much as you expected.

In article , VanguardLH
wrote:

I just got FIOS yesterday, so today I ran speed tests.

Verizons says I have 3/1 (no decimal places iirc) and that's what I
paid for. But after I ran it a few times with Ookla, it said it was
4.1 down / 1.9 up.

4.1 and 1.9 are just numerical values with no magnitude, like Kbps,
Mbps, or Gbps.

no need, since it can only be megabit.

That's like saying you car is moving at 50. 50 what?
Could be miles per hour, kilometers per hour, knots (if you have one of
those amphibious cars), or some other magnitude. 3/1 could be 3 Kbps up
and 1 Kbps down, but that's worse than dial-up service.

yep

Could be 3 Mbps
down and 1 Mbps up, but that would be damn slow even for copper
connects.

nope. that's typical for copper.

While fiber only runs up to the service entry to your home or
maybe just to the doghouse station in the neighborhood, the rest is
still copper.

fios is fibre to the home (possibly with a short run of coax between
the ont and modem/router, depending on the install), at which point
it's cat5e and/or wifi inside the home.

I would think FIOS should give 3 Gbps down and 1 Gbps up.

nope. fios is currently 100 mbit to 1 gigabit, limited by the ont.

What /speeds/ is the 3/1 plan supposed to give you?

3 mbit down, 1 mbit up, which are dsl speeds.

I typically run a speed test about 5 times. That's to try to get rid of
any caching effects on the first one or two runs and account for any
variation in how busy are the nodes (hops) in the route between me and
the test server.

Using your own ISP's speed test server is more accurate then using
something outside their network (e.g., speedtest.net by Ookla). Your
ISP only has responsibility for trying to provide you with the service
tier for which you pay them, not for anything outside their network.

nope. people are paying for internet connectivity to the world, not
just the isp's own servers.

also, their speed tests results are usually inflated to make the
service look better than it actually is.

it's best to use something impartial and to make sure connections to
the various sites people want to use work as expected.

micky wrote:
I just got FIOS yesterday, so today I ran speed tests.

Verizons says I have 3/1 (no decimal places iirc) and that's what I paid
for. But after I ran it a few times with Ookla, it said it was 4.1 down
/ 1.9 up. My friend told me that, like a car on a cold morning, it's
because after you run it a few times you get the wires warmed up and
they work better. How many times should I run it to get the best
possible test?

I'm curious what they're making you pay for this magnanimous
offering of 3Mbit down and 1Mbit up :-)

That kind of rate is what you'd run over dental fiber
(TOSLink audio is 6mbit/sec).

Your architecture could look like this. This is how my DSL is done.

CO --- fiber ---- concentrator at ----- 500ft twisted pair --- DSL 3/1
street corner copper line ADSL modem/router

That's probably about as close to FIOS as you
got, is it's 500 feet from you.

Your 1.9 result though, is anomalous, and implies there's
something wrong with my picture. You could do it this way,
but it would mean a different modem. VDSL and VDSL2 makes
it possible to easily do 10Mbit up, giving you a way to
get the 1.9Mbit number. VDSL uses more spectrum on the
twisted pair. They don't generally put the VDSL in the
garage, and if you received a new modem, you'd know it.
So this probably isn't it.

CO --- fiber ---- concentrator at ----- 500ft twisted pair --- ???
street corner copper line VDSL modem/router

While you can do a 32 way split with PON, it hardly seems
worthwhile for that low a tariffed rate. But maybe it really
is fiber ???

CO --- fiber ---- concentrator at ------------- fiber --------- Fiber optic unit
street corner (PON) in garage

That's a hell of an expensive install, for the tariff involved.
They want "$200 a month suckers" for that kind of fiber connection.

A Passive Optical Network, is one emitter and 32 fibers that have been
"welded" to the emitter. You can take a bundle of 32 glass fibers,
heat it up until it melts, and that's how you distribute copies
of the same optical signal, on 32 downstream fibers. The fiber runs
at a tremendous rate, but is time shared so everyone gets a slice.
Presumably the upstream is on a different wavelength, and uses CSMA
or something.

OK, in the example here, the upstream is done with TDMA.
That makes more sense.

https://en.wikipedia.org/wiki/Passive_optical_network

If you have a PON product, the demarc should have an RJ45
Ethernet connector, rather than you being offered an RJ11
(twisted pair) or coax cable (DOCSIS). They could do the
install in your garage, so you'd be (less) aware of any
new equipment used. It probably isn't legal (fire insurance)
to be running Ethernet cables outside the house to the
equipment, so if new equipment was brought to your
place, it's probably some place sheltered.

Paul

On Wed, 13 Nov 2019 13:52:45 -0600, VanguardLH wrote:

"wires warmed up". Did you actually believe that?

Loses due to resistance in optical or electrical cables produce a tiny
amount of heat. Cables with zero resistance do not exist.

Steve

--
http://www.npsnn.com

Stephen Wolstenholme wrote:
On Wed, 13 Nov 2019 13:52:45 -0600, VanguardLH wrote:

"wires warmed up". Did you actually believe that?

Loses due to resistance in optical or electrical cables produce a tiny
amount of heat. Cables with zero resistance do not exist.

Steve


Actually, they do.

http://www.supraconductivite.fr/en/i...sistance-supra

Paul

On 2019-11-14 6:45 a.m., Paul wrote:
Stephen Wolstenholme wrote:
On Wed, 13 Nov 2019 13:52:45 -0600, VanguardLH wrote:

"wires warmed up".Â* Did you actually believe that?

Loses due to resistance in optical or electrical cables produce a tiny
amount of heat. Cables with zero resistance do not exist.
Steve


Actually, they do.

http://www.supraconductivite.fr/en/i...sistance-supra

Â*Â* Paul

yep, that's what makes the Large Hadron Collider work.

Rene

Stephen Wolstenholme wrote:

VanguardLH wrote:

"wires warmed up". Did you actually believe that?

Loses due to resistance in optical or electrical cables produce a tiny
amount of heat. Cables with zero resistance do not exist.

You really think the small increase in resistance in wires that are
warmed up from use is going to be measurable, especially over a network
where you are hopping between hosts? The contention in Ethernet and
just one packet having to get retransmitted wipes out any of the speed
loss from a warmed Cat5 cable. There is no zero delay in getting your
packet through a host. You thought the OP had a direct connection from
the RJ45 port on his computer to the server doing the actual test?

Ookla doesn't even run their own speed test server. They connect you to
one operated by someone else. Go to speedtest.net and you'll see to
which one they select, but you can change it. Even if you happen to get
the same server in a series of tests, you don't know if you're going
through a frontend or boundary host to a server farm and the routing
inside that farm to which host happens to get selected for your test.
Hell, you don't even know if the route is the same (which hosts you
bounce through) when using the same server. Not sure how they decide
which server to use for your test, but I've selected others that gave
better results than their choice, like picking one in far away in
Chicago because it's on the same backbone network as my ISP being slower
than selecting a local university. Hell, the latency through each node
(host) in a route can change in an instant since that node would be
useless if it were quiescent. Each node and the target server have
varying levels of busy, they have limited resources, you wait your turn,
and the wait varies. Any change in resistance due to temperature
variation is severely swamped by how networking and routing works.

Ookla now lets you choose between multiple and single connections, but
they don't describe the difference. There is no way they can actually
give you a single connection from the RJ-45 port of your computer to
whatever server is selected. You still have to bounce through hosts
just to reach them, and the same for you to whatever server is used.
I'm guessing this option decides if you have a more direct route (which
is still not a single connection between you and server with no hops in
the route between) and multi-connection mode is Ookla using multiple
nodes to connect to the server. The client, through whatever routes, is
connecting to a broker (Ookla) that attempts to use multiple paths to
multiple nodes to connect to the target host. Something like:

https://www.developerxively.com/docs/connection-modes

Ookla's speedtest site defaults to multi-connections mode, so you have
even less a clue which routes (not just one route) you got to the target
server. Ookla doesn't even show you the routing for a test to know if
it is the same for subsequent tests. Of course, even if you happen to
get the same route for each test, and other than your endpoint in the
route(s), you're not the only user of the hosts in the route(s) between
you and the speed test server.

"Speedtest offers a network of over 8,000 hosted servers around the
globe" (https://www.speedtest.net/about/knowledge/faq#server). Nope,
they don't own, operate, or manage any of them. I've have never seen
the speedtest.net site offer a test server that they owned. Click on
Change Server to get a list of which are closest to you (regarding
delay, not physical distance, since a host farther away might have less
delay than a physically closer one). None of those listed are owned by
Ookla. Their statement is a lie. Having permission to use a host is
not the same as owning that host. There are tons of sites that are
webhosted, but they don't own any of the hardware (and why those site
owners choose that route to eliminate them having to do all that work
and expense).
Note: I was hunting around their site trying to find out what was the
difference between "multi" and "single" connections mode, and then read
their above statement.

https://support.speedtest.net/hc/en-...Speedtest-net-
Those are instructions on how to register YOUR host in Ookla's access
list. You run their software on your hardware. This is how Ookla gets
away with "Ookla offers a network of 8000 servers". They don't own any
of them. Again, very similar to webhosting: someone lets you deliver
your web pages for your site content from someone else's hardware.

We don't even know if the FIOS service went to the service entry point
on the OP's home (and copper from there) or to one of those doghouses
along the street and copper from there. We don't know how the OP
connected his computer to his router/cable modem. Maybe he was using a
wi-fi connection instead of Ethernet. We don't even know what
temperatures were involved for his equipment, his home, outside his
home, along the wires or fiber, at the nodes in his route to the server,
or at the server, or how all those temperatures flucuated over what
range.

Someone telling the OP that letting their house wiring warm up to get
faster speed is like saying you spitting into the ocean is going to
raise the sea level or would affect the tides. Sure, your wad is going
to affect the tide based on physics, but is anyone really going to
notice the tide got slowed or sped up (depending on which way into the
tide you spit)?

On 11/14/19 8:51 AM, Rene Lamontagne wrote:
On 2019-11-14 6:45 a.m., Paul wrote:
Stephen Wolstenholme wrote:
On Wed, 13 Nov 2019 13:52:45 -0600, VanguardLH wrote:

"wires warmed up".Â* Did you actually believe that?

Loses due to resistance in optical or electrical cables produce a tiny
amount of heat. Cables with zero resistance do not exist.
Steve


Actually, they do.

http://www.supraconductivite.fr/en/i...sistance-supra

Â*Â*Â* Paul

yep, that's what makes the Large Hadron Collider work.

Rene

We need something that is superconductive at 48 degrees F.

--
41 days until the winter celebration (Wed, Dec 25, 2019 12:00:00 AM for
1 day).

"Irreligion: The principal one of the great faiths of the world." --
Ambrose Bierce

On Thu, 14 Nov 2019 12:41:01 -0600, VanguardLH wrote:

Ookla doesn't even run their own speed test server. They connect you to
one operated by someone else. Go to speedtest.net and you'll see to
which one they select, but you can change it. Even if you happen to get
the same server in a series of tests, you don't know if you're going
through a frontend or boundary host to a server farm and the routing
inside that farm to which host happens to get selected for your test.

It's mostly irrelevant. You're trying to measure end to end throughput, so
you don't care about the infrastructure as long as the test is valid, and
you can determine whether it is by repeating the test multiple times, with
multiple remote hosts.

Hell, you don't even know if the route is the same (which hosts you
bounce through) when using the same server. Not sure how they decide
which server to use for your test, but I've selected others that gave
better results than their choice, like picking one in far away in
Chicago because it's on the same backbone network as my ISP being slower
than selecting a local university. Hell, the latency through each node
(host) in a route can change in an instant since that node would be
useless if it were quiescent. Each node and the target server have
varying levels of busy, they have limited resources, you wait your turn,
and the wait varies. Any change in resistance due to temperature
variation is severely swamped by how networking and routing works.

As above, the route taken is mostly irrelevant. The Internet is designed in
such a way that traffic finds its way to its destination, one hop at a
time, and for the most part neither endpoint has much to say about it. You
just want the test(s) to be valid, and you can help by repeating the test
multiple times, with multiple remote hosts.

Ookla now lets you choose between multiple and single connections, but
they don't describe the difference. There is no way they can actually
give you a single connection from the RJ-45 port of your computer to
whatever server is selected. You still have to bounce through hosts
just to reach them, and the same for you to whatever server is used.

Multiple versus single connections is just like it sounds. Many times,
single connections are rate limited in one way or another, so by opening
multiple *concurrent* connections, you might see an improved overall
throughput. Or, you might not, thus the option to try both ways.

If you've ever used an HTTP download manager, a Usenet binary downloader,
or a Bittorrent client, you're already aware that you can open multiple
concurrent connections, which tends to greatly increase the throughput
performance.

I'm guessing this option decides if you have a more direct route (which
is still not a single connection between you and server with no hops in
the route between) and multi-connection mode is Ookla using multiple
nodes to connect to the server. The client, through whatever routes, is

No, it's nothing to do with the number of hops.

connecting to a broker (Ookla) that attempts to use multiple paths to
multiple nodes to connect to the target host. Something like:

https://www.developerxively.com/docs/connection-modes

Ookla's speedtest site defaults to multi-connections mode, so you have
even less a clue which routes (not just one route) you got to the target
server.

It's the Internet. By definition, you don't know the route that traffic
takes to get to you, and no two packets necessarily need to take the same
path. Take OSPF, for example (Open Shortest Path First). As each packet
hits a transit router, the router makes a decision as to which way to send
that packet to get it one hop closer to its destination. When the next
packet arrives, a whole new decision is made, and so on.

Ookla doesn't even show you the routing for a test to know if
it is the same for subsequent tests. Of course, even if you happen to
get the same route for each test, and other than your endpoint in the
route(s), you're not the only user of the hosts in the route(s) between
you and the speed test server.

As above, the route is irrelevant. The aim is to measure end to end
throughput.

"Speedtest offers a network of over 8,000 hosted servers around the
globe" (https://www.speedtest.net/about/knowledge/faq#server). Nope,
they don't own, operate, or manage any of them. I've have never seen
the speedtest.net site offer a test server that they owned. Click on

Is it significant that they own, or don't own, the servers? Again, if
you're running a speed test, you just want to measure the throughput
performance. Server ownership isn't part of the equation.

Change Server to get a list of which are closest to you (regarding
delay, not physical distance, since a host farther away might have less
delay than a physically closer one). None of those listed are owned by
Ookla. Their statement is a lie. Having permission to use a host is
not the same as owning that host. There are tons of sites that are
webhosted, but they don't own any of the hardware (and why those site
owners choose that route to eliminate them having to do all that work
and expense).
Note: I was hunting around their site trying to find out what was the
difference between "multi" and "single" connections mode, and then read
their above statement.

https://support.speedtest.net/hc/en-...Speedtest-net-
Those are instructions on how to register YOUR host in Ookla's access
list. You run their software on your hardware. This is how Ookla gets
away with "Ookla offers a network of 8000 servers". They don't own any
of them. Again, very similar to webhosting: someone lets you deliver
your web pages for your site content from someone else's hardware.

I don't see any issues with the claim that "Ookla offers a network of 8000
servers". I can't vouch for the number being claimed, but I have no issues
regarding ownership of the equipment. I don't think it matters at all.

snip