Download speeds

I realize that there will be bottlenecks when doing downloads.

My best dl speed is about 400 kb/sec under Linux.

(If you love Windows, do not ask what my top speed is under XP. :-)

Even at off peak times.

Why is it not even close to what Speedtest shows ?

Andy



Retrieving speedtest.net configuration...
Retrieving speedtest.net server list...
Testing from Verizon Internet Services (71.97.192.187)...
Selecting best server based on latency...
Hosted by T-Mobile (Houston, TX) [36.56 km]: 58.667 ms
Testing download speed........................................
Download: 2.85 Mbit/s

Andy wrote:
I realize that there will be bottlenecks when doing downloads.

My best dl speed is about 400 kb/sec under Linux.

(If you love Windows, do not ask what my top speed is under XP. :-)

Even at off peak times.

Why is it not even close to what Speedtest shows ?

Andy



Retrieving speedtest.net configuration...
Retrieving speedtest.net server list...
Testing from Verizon Internet Services (71.97.192.187)...
Selecting best server based on latency...
Hosted by T-Mobile (Houston, TX) [36.56 km]: 58.667 ms
Testing download speed........................................
Download: 2.85 Mbit/s

The speed of the server end matters. The end to end
connection is limited to the "weakest link".

That's one of the first things you notice, after
an Internet speed upgrade - hardly any of the
servers you use, can keep up. It's kinda deflating
to discover that.

Paul

On Thursday, April 7, 2016 at 6:10:37 PM UTC-5, Paul wrote:
Andy wrote:
I realize that there will be bottlenecks when doing downloads.

My best dl speed is about 400 kb/sec under Linux.

(If you love Windows, do not ask what my top speed is under XP. :-)

Even at off peak times.

Why is it not even close to what Speedtest shows ?

Andy



Retrieving speedtest.net configuration...
Retrieving speedtest.net server list...
Testing from Verizon Internet Services (71.97.192.187)...
Selecting best server based on latency...
Hosted by T-Mobile (Houston, TX) [36.56 km]: 58.667 ms
Testing download speed........................................
Download: 2.85 Mbit/s

The speed of the server end matters. The end to end
connection is limited to the "weakest link".

That's one of the first things you notice, after
an Internet speed upgrade - hardly any of the
servers you use, can keep up. It's kinda deflating
to discover that.

Paul

So, are you saying that my internet provider is the weak link?

If, so I will consider going with a lower speed and save some dough. :-)

Andy

On Thursday, April 7, 2016 at 6:10:37 PM UTC-5, Paul wrote:
Andy wrote:
I realize that there will be bottlenecks when doing downloads.

My best dl speed is about 400 kb/sec under Linux.

(If you love Windows, do not ask what my top speed is under XP. :-)

Even at off peak times.

Why is it not even close to what Speedtest shows ?

Andy



Retrieving speedtest.net configuration...
Retrieving speedtest.net server list...
Testing from Verizon Internet Services (71.97.192.187)...
Selecting best server based on latency...
Hosted by T-Mobile (Houston, TX) [36.56 km]: 58.667 ms
Testing download speed........................................
Download: 2.85 Mbit/s

The speed of the server end matters. The end to end
connection is limited to the "weakest link".

That's one of the first things you notice, after
an Internet speed upgrade - hardly any of the
servers you use, can keep up. It's kinda deflating
to discover that.

Paul

It is not deflating, it is a rip-off.

Paying more and not getting much for it.

Andy

Andy wrote:
On Thursday, April 7, 2016 at 6:10:37 PM UTC-5, Paul wrote:
Andy wrote:
I realize that there will be bottlenecks when doing downloads.

My best dl speed is about 400 kb/sec under Linux.

(If you love Windows, do not ask what my top speed is under XP. :-)

Even at off peak times.

Why is it not even close to what Speedtest shows ?

Andy



Retrieving speedtest.net configuration...
Retrieving speedtest.net server list...
Testing from Verizon Internet Services (71.97.192.187)...
Selecting best server based on latency...
Hosted by T-Mobile (Houston, TX) [36.56 km]: 58.667 ms
Testing download speed........................................
Download: 2.85 Mbit/s
The speed of the server end matters. The end to end
connection is limited to the "weakest link".

That's one of the first things you notice, after
an Internet speed upgrade - hardly any of the
servers you use, can keep up. It's kinda deflating
to discover that.

Paul

It is not deflating, it is a rip-off.

Paying more and not getting much for it.

Andy

Well, I can relate my ADSL experience.

I was using a lower tier plan. There are
two caps involved. You pay for an "up to"
rate in megabits/sec. But the actual setup
is set to a second, lower cap value (you don't
know what this is in advance). It is supposed
to be set by measured SNR, but lazy staff
just dial it down where they know it'll be safe.

The end result, is you get about half what
you're paying for.

Now, the other interesting thing (their explanation
for why it's not as bad as it looks), is the PPPOE
encapsulation has a bit of overhead. And the
"up to" rate includes the overhead. That's why
your (in-computer) measured rate, and their measured
rate on their end, differs by the percentage
overhead for PPPOE (point to point protocol over
etherhet). A second way of doing it, is PPPOA or
point to point protocol over 53-byte asynchronous
transfer mode cells. If your service has a
VPI:VCI, then that means it uses PPPOA as
far as I know. PPPOE is probably more
like HDLC protocol.

Now, when I got my ADSL upgrade (same price but
reduced gigabytes per month), two things
happened. The link rate was 3 times higher,
but for some reason, they started measuring
the performance in the units I use on my end
of the link (so I'm no longer getting ripped
off for the overhead factor). So now the
performance *exactly matches* the contract rate.
How weird is that ? I know the SNR on the short hop to
my street corner is going to be excellent,
but it's still a shock to see that "poor piece
of wire" suddenly perform like a champ. It means
the 50% rate I was getting on my lower tier,
was a real ripoff. Based on past experience,
I figured at best, I'd get 80% of what I
was paying for. But on the first try, it
went 100%.

*******

Your 400 rate, you should first verify
your units. 2.85Mbit/sec divided by 8 bits
per byte, is 356KB/sec (the units in your
browser download dialog box). So the 400
measurement you mention, is likely
to be 400KB/sec rather than Kbits/sec.
In other words, when the browser says
"about 400", that is actual 3200Kbit/sec
or in the vicinity of the Speedtest value.

Still, on that tier, I expect you're paying for
5Mbit/sec and getting 2.85Mbit/sec. The actual
rate on the link (includes overhead), might
be on the order of 2.85*1.15 = 3.28Mbit/sec.
And then your "ripoff" level is 3.28/5.0 = 66%
of the contract rate.

You could raise a trouble ticket, and have the
line checked. Now, if you happen to be 18000 feet
from the CO, as opposed to 500 feet from the
concentrator on the street corner, that makes
all the difference in the world. If you are 500 feet
from a new box put on your street corner, then
you should be getting the whole contract rate
(5Mbit/sec on low tier). If you're 18,000 feet
away from the terminating equipment, then there's
no way you'll get the full rate. Urban installs
use concentrators at the corner. Rural installs
(how mine used to be set up before the box
arrived), they run the cable all the way back
to the CO or vault.

There is an application called DMT, which reads out
the ADSL modem frequency buckets, and it also
gives the SNR margin. My ISP would accept information
of that sort, for the purpose of raising the trouble
ticket. If there is buckets of SNR margin (at least
6dB), the "working cap" on the line can be raised.
But of course, the working cap cannot be higher
than the contract rate.

What's neat about how ADSL works, is instead
of allowing the line rate to rise until errors
start to show up, instead the stupid thing works
on a static allocation. For example, the tech who
did a truck roll to my place, he had a gadget on
the line while he was in the house, and it was
reading out the parameters while we talked. And
that guy could see right away there were buckets
or margin. As he didn't have to find another
line pair for my install. The SNR can be measured
from either end of the line, and his display showed
what the concentrator could see.

They measured all our line pairs, a couple summers
ago. That was so they could pick out the best ones,
for the triple play customers paying $100+ a month.
There's only two or three of those people
on my street. I could tell, because the truck stays
at a place like that for practically a whole day,
while the VDSL box is installed in the garage.
My service, on the other hand, still uses a
regular ADSL2+ modem and is a bit more modest.
No Internet TV.

Paul

On Thursday, April 7, 2016 at 9:55:00 PM UTC-5, Paul wrote:
Andy wrote:
On Thursday, April 7, 2016 at 6:10:37 PM UTC-5, Paul wrote:
Andy wrote:
I realize that there will be bottlenecks when doing downloads.

My best dl speed is about 400 kb/sec under Linux.

(If you love Windows, do not ask what my top speed is under XP. :-)

Even at off peak times.

Why is it not even close to what Speedtest shows ?

Andy



Retrieving speedtest.net configuration...
Retrieving speedtest.net server list...
Testing from Verizon Internet Services (71.97.192.187)...
Selecting best server based on latency...
Hosted by T-Mobile (Houston, TX) [36.56 km]: 58.667 ms
Testing download speed........................................
Download: 2.85 Mbit/s
The speed of the server end matters. The end to end
connection is limited to the "weakest link".

That's one of the first things you notice, after
an Internet speed upgrade - hardly any of the
servers you use, can keep up. It's kinda deflating
to discover that.

Paul

It is not deflating, it is a rip-off.

Paying more and not getting much for it.

Andy

Well, I can relate my ADSL experience.

I was using a lower tier plan. There are
two caps involved. You pay for an "up to"
rate in megabits/sec. But the actual setup
is set to a second, lower cap value (you don't
know what this is in advance). It is supposed
to be set by measured SNR, but lazy staff
just dial it down where they know it'll be safe.

The end result, is you get about half what
you're paying for.

Now, the other interesting thing (their explanation
for why it's not as bad as it looks), is the PPPOE
encapsulation has a bit of overhead. And the
"up to" rate includes the overhead. That's why
your (in-computer) measured rate, and their measured
rate on their end, differs by the percentage
overhead for PPPOE (point to point protocol over
etherhet). A second way of doing it, is PPPOA or
point to point protocol over 53-byte asynchronous
transfer mode cells. If your service has a
VPI:VCI, then that means it uses PPPOA as
far as I know. PPPOE is probably more
like HDLC protocol.

Now, when I got my ADSL upgrade (same price but
reduced gigabytes per month), two things
happened. The link rate was 3 times higher,
but for some reason, they started measuring
the performance in the units I use on my end
of the link (so I'm no longer getting ripped
off for the overhead factor). So now the
performance *exactly matches* the contract rate.
How weird is that ? I know the SNR on the short hop to
my street corner is going to be excellent,
but it's still a shock to see that "poor piece
of wire" suddenly perform like a champ. It means
the 50% rate I was getting on my lower tier,
was a real ripoff. Based on past experience,
I figured at best, I'd get 80% of what I
was paying for. But on the first try, it
went 100%.

*******

Your 400 rate, you should first verify
your units. 2.85Mbit/sec divided by 8 bits
per byte, is 356KB/sec (the units in your
browser download dialog box). So the 400
measurement you mention, is likely
to be 400KB/sec rather than Kbits/sec.
In other words, when the browser says
"about 400", that is actual 3200Kbit/sec
or in the vicinity of the Speedtest value.

Still, on that tier, I expect you're paying for
5Mbit/sec and getting 2.85Mbit/sec. The actual
rate on the link (includes overhead), might
be on the order of 2.85*1.15 = 3.28Mbit/sec.
And then your "ripoff" level is 3.28/5.0 = 66%
of the contract rate.

You could raise a trouble ticket, and have the
line checked. Now, if you happen to be 18000 feet
from the CO, as opposed to 500 feet from the
concentrator on the street corner, that makes
all the difference in the world. If you are 500 feet
from a new box put on your street corner, then
you should be getting the whole contract rate
(5Mbit/sec on low tier). If you're 18,000 feet
away from the terminating equipment, then there's
no way you'll get the full rate. Urban installs
use concentrators at the corner. Rural installs
(how mine used to be set up before the box
arrived), they run the cable all the way back
to the CO or vault.

There is an application called DMT, which reads out
the ADSL modem frequency buckets, and it also
gives the SNR margin. My ISP would accept information
of that sort, for the purpose of raising the trouble
ticket. If there is buckets of SNR margin (at least
6dB), the "working cap" on the line can be raised.
But of course, the working cap cannot be higher
than the contract rate.

What's neat about how ADSL works, is instead
of allowing the line rate to rise until errors
start to show up, instead the stupid thing works
on a static allocation. For example, the tech who
did a truck roll to my place, he had a gadget on
the line while he was in the house, and it was
reading out the parameters while we talked. And
that guy could see right away there were buckets
or margin. As he didn't have to find another
line pair for my install. The SNR can be measured
from either end of the line, and his display showed
what the concentrator could see.

They measured all our line pairs, a couple summers
ago. That was so they could pick out the best ones,
for the triple play customers paying $100+ a month.
There's only two or three of those people
on my street. I could tell, because the truck stays
at a place like that for practically a whole day,
while the VDSL box is installed in the garage.
My service, on the other hand, still uses a
regular ADSL2+ modem and is a bit more modest.
No Internet TV.

Paul

Interesting experience.

Andy

On Friday, April 8, 2016 at 8:10:56 AM UTC-5, Andy wrote:
On Thursday, April 7, 2016 at 9:55:00 PM UTC-5, Paul wrote:
Andy wrote:
On Thursday, April 7, 2016 at 6:10:37 PM UTC-5, Paul wrote:
Andy wrote:
I realize that there will be bottlenecks when doing downloads.

My best dl speed is about 400 kb/sec under Linux.

(If you love Windows, do not ask what my top speed is under XP. :-)

Even at off peak times.

Why is it not even close to what Speedtest shows ?

Andy



Retrieving speedtest.net configuration...
Retrieving speedtest.net server list...
Testing from Verizon Internet Services (71.97.192.187)...
Selecting best server based on latency...
Hosted by T-Mobile (Houston, TX) [36.56 km]: 58.667 ms
Testing download speed........................................
Download: 2.85 Mbit/s
The speed of the server end matters. The end to end
connection is limited to the "weakest link".

That's one of the first things you notice, after
an Internet speed upgrade - hardly any of the
servers you use, can keep up. It's kinda deflating
to discover that.

Paul

It is not deflating, it is a rip-off.

Paying more and not getting much for it.

Andy

Well, I can relate my ADSL experience.

I was using a lower tier plan. There are
two caps involved. You pay for an "up to"
rate in megabits/sec. But the actual setup
is set to a second, lower cap value (you don't
know what this is in advance). It is supposed
to be set by measured SNR, but lazy staff
just dial it down where they know it'll be safe.

The end result, is you get about half what
you're paying for.

Now, the other interesting thing (their explanation
for why it's not as bad as it looks), is the PPPOE
encapsulation has a bit of overhead. And the
"up to" rate includes the overhead. That's why
your (in-computer) measured rate, and their measured
rate on their end, differs by the percentage
overhead for PPPOE (point to point protocol over
etherhet). A second way of doing it, is PPPOA or
point to point protocol over 53-byte asynchronous
transfer mode cells. If your service has a
VPI:VCI, then that means it uses PPPOA as
far as I know. PPPOE is probably more
like HDLC protocol.

Now, when I got my ADSL upgrade (same price but
reduced gigabytes per month), two things
happened. The link rate was 3 times higher,
but for some reason, they started measuring
the performance in the units I use on my end
of the link (so I'm no longer getting ripped
off for the overhead factor). So now the
performance *exactly matches* the contract rate.
How weird is that ? I know the SNR on the short hop to
my street corner is going to be excellent,
but it's still a shock to see that "poor piece
of wire" suddenly perform like a champ. It means
the 50% rate I was getting on my lower tier,
was a real ripoff. Based on past experience,
I figured at best, I'd get 80% of what I
was paying for. But on the first try, it
went 100%.

*******

Your 400 rate, you should first verify
your units. 2.85Mbit/sec divided by 8 bits
per byte, is 356KB/sec (the units in your
browser download dialog box). So the 400
measurement you mention, is likely
to be 400KB/sec rather than Kbits/sec.
In other words, when the browser says
"about 400", that is actual 3200Kbit/sec
or in the vicinity of the Speedtest value.

Still, on that tier, I expect you're paying for
5Mbit/sec and getting 2.85Mbit/sec. The actual
rate on the link (includes overhead), might
be on the order of 2.85*1.15 = 3.28Mbit/sec.
And then your "ripoff" level is 3.28/5.0 = 66%
of the contract rate.

You could raise a trouble ticket, and have the
line checked. Now, if you happen to be 18000 feet
from the CO, as opposed to 500 feet from the
concentrator on the street corner, that makes
all the difference in the world. If you are 500 feet
from a new box put on your street corner, then
you should be getting the whole contract rate
(5Mbit/sec on low tier). If you're 18,000 feet
away from the terminating equipment, then there's
no way you'll get the full rate. Urban installs
use concentrators at the corner. Rural installs
(how mine used to be set up before the box
arrived), they run the cable all the way back
to the CO or vault.

There is an application called DMT, which reads out
the ADSL modem frequency buckets, and it also
gives the SNR margin. My ISP would accept information
of that sort, for the purpose of raising the trouble
ticket. If there is buckets of SNR margin (at least
6dB), the "working cap" on the line can be raised.
But of course, the working cap cannot be higher
than the contract rate.

What's neat about how ADSL works, is instead
of allowing the line rate to rise until errors
start to show up, instead the stupid thing works
on a static allocation. For example, the tech who
did a truck roll to my place, he had a gadget on
the line while he was in the house, and it was
reading out the parameters while we talked. And
that guy could see right away there were buckets
or margin. As he didn't have to find another
line pair for my install. The SNR can be measured
from either end of the line, and his display showed
what the concentrator could see.

They measured all our line pairs, a couple summers
ago. That was so they could pick out the best ones,
for the triple play customers paying $100+ a month.
There's only two or three of those people
on my street. I could tell, because the truck stays
at a place like that for practically a whole day,
while the VDSL box is installed in the garage.
My service, on the other hand, still uses a
regular ADSL2+ modem and is a bit more modest.
No Internet TV.

Paul

Interesting experience.

Andy

I disconneted Netflix, but there was no change in dl speed.

Frontier bought out my landline a year ago.

They still are not setup to let me pay my last bill.

Quite unprofessional.

And govt. in their wisdom broke up Ma Bell.

Supposed to help the consumer.

NOT !!

Now each area has only one supplier for landlines. ??

I wish I could use AT&T. MUCH better company.

Andy