Clearing up some misconceptions.

Good morning.

I subscribed to this group yesterday because I was having a specific
problem with my setup - see elsewhere - in the hope that here I might
find some people with more knowledge than I had. Well so far one or two
have responded and for that I thank them.

However on scanning through the remainder of the threads, it has become
fairly obvious that the vast majority of people here know a lot LESS
than I do.

In particular a recurring theme is 'will I get better performance if I
change my ISP' and since the answer to this in most cases is a
resounding NO. I thought a dump of knowledge might be a good idea.

I spent between about 1987 and 2000 as technical support, technical
director and finally managing director of two companies specializing in
WAN networking, so although my detailed knowledge is rusty and a little
out of date the background is good. In 1999 I sold my company and in
2000 I left it as frankly I was totally exhausted..but you don't want to
know about that. Let's say since then I haven't wanted to look a router
in the face unless I had to - I am basically retired..sort of.
Now to business.

LOCAL LOOP
==========

Unless you are getting connectivity via wireless or cable, or possibly
fiber, your broadband is delivered down pair of copper wires, and there
is a 99% likelihood that whoever your ISP is, it's the same pair of wires.

Moreover, unless you live in a metropolitan environment, what those
wires plug into at the exchange is going to be the same piece of kit - a
DSLAM - supplied, managed and operated by BT. It IS possible for larger
suppliers to have their own DSLAMS in BTs exchanges, but it is less
common than they like you to think.

THOSE WIRES AND THAT DSLAM LIMIT THE MAX SPEED YOU CAN GET FROM THE
BROADBAND. *If* your speeds are low because of deficiencies *there*,
changing ISPs will make zero difference to the speed you get.

Despite what the ISP's would have you believe.

I myself am with Clara net, and the product they are selling is EXACTLY
technically equivalent to the BT Max product,which is unsurprising since
it runs over BT kit.

Now other things than the local loop can *REDUCE* the *FINAL* speed you
get, overall, in data transfers, *but if you are only reliably
connecting at - say - 3MBPS - changing ISP will not net you better
performance*.

In the ultimate analysis the local loop remains the limiting factor.

BACKHAUL
========
This egregious term is used to describe the process by which the local
exchange multiplexes and agglomerates the various connections into the
DSLAMS into streams that are fed to the various ISPS. Once again,these
are not separate physical feeds: They are not even strictly separate
logical feeds. BT operates an ATM network - a sort of private low level
internet in its own right - to take data packets from the DSLAMs,
interleave them down a multiplicity of physical channels (fibre,
microwave and god knows what else) and finally deliver them to the ISPs
ATM interfaces somewhere in a vast machine room in London. (or
occasionally Manchester IIRC). The point of describing this is to show
you that by and large, although I think BT offer graded backhaul at
different qualities of service and so called contention ratios, all the
ISP'S tend to use *their* ATM backhaul to get the packets from the local
exchange to the ISP, and once again, unless you pay for a service with a
lower contention ratio than your previous one, the chances are that this
part of the overall link will not change in the slightest if you change ISP.

I'll get to what happens afterwards later, but let me again make the
point. NO MATTER WHICH ISP YOU CHOOSE, UP TO THE ATM INTERFACE TO THEIR
NETWORK, THE WIRES, CABLES, DSLAMS AND LINKS ARE PRECISELY THE SAME.
Unless they are rally cheapskate on their kit and make a pigs ear of the
ATM interface to BT, there won't be one jot of difference who you go to.
You are just buying the same rebranded BT service as you were before.

ISP
====
At Telehouse or wherever..you actually hit the ISP's network proper.
nearly all the ISPS have ther main centers in one of half a dozen
buildings in london, where their racks sit close to or even side by side
with their competitors , and pass data to them down usually high speed
fiber connections. That's where all the servers are too ..mail DNS, web
etc etc. All sitting in one building with very cheap and very fast
connections between them. There will be interconnects - gigabit fiber -
between buildings - again typically run by BT but other players also
have fibre running between these. The net result of this is that by and
large, all UK ISPS can talk to each other pretty much flat out with few
speed restrictions. So to access a UK based server via your ISP should
make little difference WHICH ISP you are with. Last night I tried speed
tests on every googleable speed tester I could find, and apart from one
notable exception whose name I have truly forgotten, all showed me a
virtually identical speed.

The ONLY differences you MAY see finally between ISP's in speed terms,
are in international bandwidth. Most smaller ISPs buy that off larger
ISPs,and not a lot of it. Larger ones have their own, but depending on
congestion and who they peer with elsewhere in the world, performance
may vary markedly from one ISP to another. But unless you are accessing
foreign web sites etc, this won't be of huge interest to you. I am not
sure what the stats are today, but when I was last in a position to
know, international traffic was typically between 10% and 30% of all UK
traffic.

So leaving access to servers aside, these days by and large even
international speeds are unlikely to change between ISP's. In terms of
pure connectivity, most major ISPs are by and large similar.

The economics of running an ISP are interesting. I can't exactly say
what todays mix is, but when I ran one, the biggest cost was staff -
mainly sales and support staff, with admin marketing and top techies
being about equal. In other words. what cost the money was getting
customers and keeping them. Running the network as such took very few
people. Line costs were quite high too. So international bandwidth was
always at a premium. However servers and routers were simply commodity
items. A core router cost about the same as keeping a support person for
a month, and in financial terms, amortized over say 5 years of life,
given access to debt financing, a new router cost less than a support
person for a week.

No wonder it was cisco everywhere. We could afford to pay anything to
get rid of support problems. Also, today, no wonder support is from the
'cow shed in Chupattistan'. You should understand this. Its billions
cheaper to build a good network at the ISP level than to provide people
to listen to your problems.

DEBUGGING THE PROTOCOL CHAIN
============================

Again I was shocked here to see things like 'ADSL password' being
bandied about. Since there is almost zero chance of getting quality
support from the ISPs for the reasons described above, I thought it
worthwhile to pass on such as I do know to be a possibly useful resoucre
in debugging ADSL type broadband..

First of all, ADSL is merely the protocol between your router or modem,
and the DSLAM at the exchange. It is basically a way to cram as much
data down a 19th century pair of telegraph wires as modern technology
can achieve at consumer cost levels and without melting them, or causing
aeroplanes to fall out of the sky, or using the output of a couple of
power stations to do it.

ADSL happens totally automatically without human intervention, between
your router/modem and thee DSLAM. If the router is configured to
'connect' there is a swift interchange of signals down the line, and
some calculations and line test signals go back and forward, and a
suitable rate of exchange is negotiated, and you should end up with a
more or less reliable exchange of data between your kit and the DSLAM.

In fixed rate services, this process - called 'training' - takes less
than a second. The Max rate training seems to take longer - several
seconds in fact. I'll get on to Max Rate later. Usually the modem or
router will show some kind of panel light indication that is on,
connected to a DSLAM at the other end, and that the connection is being
trained, or is established.

This is where ADSL starts `and finishes. From now on higher level
protocols take over. The data packet that goes to the DSLAM will be
forwarded over the ATM network to the ISP. If your router has the right
diagnostics, it will be able to tell you

- if the connection to the DSLAM is up - the front panel lights will
show that.
- if it can see something recognisable at the far end, and so the ATM
layer is working.

Having set up two protocols to get to the ISP, a third one now is
layered over the top. Typically that is PPP. In fact its PPPloA which I
think is PPP (point to point protocol) over ATM, PPP is nice because
unlike ADSL or ATM it contains a level of abstraction that, amongst
other things, allows the concept of user authentication. When you type
in a name and password into teh mdoem or router, its PPP that is picking
those up and using them to verify you are a valid customer to the ISP's
kit. That kit - no idea what it is, probably a huge cisco router or
similar - accepts these ppp over ATM packets, sees its you, and then
sets up an IP conversation with you layered over the PPP link. Part of
that conversation will be to pass back to you useful stuff via DHCP
telling your router what IP address to use, and where the local DNS
servers that you should use, are. Etc. Note that this DHCP is not the
same as your local LAN DHCP - its a different conversation, using the
same protocol. If e.g. you have a static IP address that does not mean
you should configure that in the router. Oh no. All that means is that
every time you log in DHCP will pass you the SAME IP address as last
time. The ISP must set that up in their user tables.

Now if you are a single PC with a modem, that's basically it `as far as
setting things up goes. TCP is layered over IP which is layered over PPP
which is transported to your ISP via ATM from the exchange, and ADSL to
you from the exchange.

If you have decent diagnostics, problems can be quickly narrowed down to
whichever of these is the problem. Lack of ADSL means the line, or the
remote DSLAM is tits up. A BT problem or inside your house.

If the ADSL is up,but the far end is not responding, its a BT ATM problem.

If the far end is up at test level, but you cannot log in, its your
ISP's kit thats tits up.

If IP is working (pings work etc) but nothing else does, is likely the
ISP has routing or server issues.

Etc.


RATE ADAPTING ADSL - BT MAX.
=============================

This deserves a bit of detail - and is very much what I have learnt in
the last few days by spending a shitload of time asking people, playing
with my setup and sifting through stuff on the web.

The first thing to note is that the support person at the other end of
your call will, if my ISP - one of the better ones - is anything to go
by, know almost nothing of any of this.

Rate adapting ADSL is as currently sold in the UK under what ever brand
it is, is an attempt to extract the best performance out of a single
copper pair that currently established technology can give. It promises
up to 8Mbps downloads and 448k uploads. If you see those magic figures,
that's what you are being sold.

Its well established technology - my 7 year old router talks it - but
its taken BT a fair time to upgrade and install the DSLAMS that talk it.
and although its been around for 2 years in the Uk or so, the support
people at the ISPs still do not understand it. Nor, I confess, do I, and
if anyone else wants to chip in and correct anything I am about to say,
feel free. We can all benefit from a better understanding.

The first thing to note is that this is bleeding edge stuff, and that
potentially makes it flakey as hell. In order to get the best out of
YOUR line it seems to do the following, at two different levels.

It has to establish a connection speed that is a compromise between
speed, and reliability. The faster the speed, the less margin for error
there is and long lines or bad lines mean less speed.

The kit at both ends negotiates a speed depending on what it finds. This
takes a long time. Wheres my 576K ADSL would reconnect in the blink of
an eye, this ADSL MAX is taking 4-8 seconds to establish a connection.
That means IF it drops its bloody noticeable.

Secondly because the ISPs all want to sell you the latest greatest and
fastest, it's set right on the limit. That means it has very little
margin of tolerance to noise. One responder to my posts says it takes as
little as a thunderstorm over France to knock his off beam.

To alleviate the flood of support calls this sort of marketing stupidity
would generate, BT have devised a cunning plan. Let's call it the
'Baldric algorithm'. For the first 10-12 days, the connection that you
negotiate will go through successively different types of modulation,
and depending on the frequency of lost connections, a putative rate for
your line will be arrived at. This is called the Maximum Stable Rate.. a
ten percent drop in this will be set as the Fault Threshold Rate. One
thing is clear, and that is that BT will NOT respond to line fault calls
if you are connecting above the Fault Threshold Rate.

Some examples. I am currently connected at 5184kbps. According to my ISP
who thoughtfully allow me to view the results of the Baldric Algorithm,.
my Maximum Stable Rate is 4544kbps and my Fault Threshold Rate is 3635kbps.

HOWEVER, in my case that connection speed is 5184kbps, and my SNR is a
piddling 4.5dB, whereas 6dB is reckoned to be the minimum for stability
over a decent period of time. BT have correctly calculated that I should
be running no faster than 4544kbps, and I wish I was. Lowering line
speed improves the SNR quite dramatically. My attenuation is 46dB. about
par for the line length - its just that the line is noisy.

A postcode test reveals that the expected max stable rate should be
about 2900kbps, so I am fairly happy that the line is OK..what is
screwing me up is the Baldric Algorithm. Or why its not being enforced
anyway.

Any comments on this gratefully received.

That's all I have time for. I hope it proves useful.

"The Natural Philosopher" <(E-Mail Removed)> wrote in message
news:(E-Mail Removed)...
> Good morning.
Lines: 319!! FFS! Get a fucking LIFE.

The Natural Philosopher wrote:

> In particular a recurring theme is 'will I get better performance if I
> change my ISP' and since the answer to this in most cases is a
> resounding NO.

I suppose you think that bandwidth throttling, port blocking and other 'traffic
management' measures make no difference to one's internet experience do you ?

You seem to be completely out of touch with reality.

Graham

In article <(E-Mail Removed)>, The Natural
Philosopher <(E-Mail Removed)> wrote:

> Good morning.
>
> I subscribed to this group yesterday because I was having a specific
> problem with my setup - see elsewhere - in the hope that here I might
> find some people with more knowledge than I had. Well so far one or two
> have responded and for that I thank them.
>
> However on scanning through the remainder of the threads, it has become
> fairly obvious that the vast majority of people here know a lot LESS
> than I do.
>
> In particular a recurring theme is 'will I get better performance if I
> change my ISP' and since the answer to this in most cases is a
> resounding NO. I thought a dump of knowledge might be a good idea.

Despite all your theoretical verbiage, the real answer is neither yes
or no - it depends on where the speed restriction is. If the
bottleneck is at the local exchange then changing ISP will have no
effect unless, as you say, by switching ISP you also move to LLU.

Practically, in my own case with nildram 'upto 8Mb' I had low average
speeds (less than 1Mbps) during weekday office-hour for many months.
Nildram told me it was my exchange and so no point in switching ISP.
Despite that I decided to switch for a variety of reasons, e.g. because
the new ISP was cheaper.

The very same day I switched to an entanet reseller on the 'upto 8Mb'I
went to average weekday office-hour speeds of over 5Mbps. In the 7
months since then that good average speed has been maintained.

As might be expected, the modem synch speed was the same with both
ISPs.

Somehow, at least in my own particular case, I find it hard to believe
that the great increase in speed just coincidentally happened on the
same day that I switched ISP. It is much more likely to be related to
some ISP-related parameter (e.g. capacity).

Kit

i agree with everyone (except OP)
i changed from pipex to entanet and my internet experience changed
dramatically, same sync speed but much faster actual speed day and night, no
port throttling so i can use the internet as i like, low pings 24/7 which
was NEVER the case before. clearly defined usage limits and penalties (with
facility to check so far)., not just some vague FUP. freephone uk based tech
support if i need it, not £1/min
much as the technology to the isp is the same, from there on it's totally
different.
a simple analogy would be driving to ireland from uk, everyone would use the
same road, but once at the pier choosing fast ferry or slow makes a major
difference, both are boats, but one performs much better than the other.

Kit wrote:
> In article <(E-Mail Removed)>, The Natural
> Philosopher <(E-Mail Removed)> wrote:
>
>> Good morning.
>>
>> I subscribed to this group yesterday because I was having a specific
>> problem with my setup - see elsewhere - in the hope that here I might
>> find some people with more knowledge than I had. Well so far one or two
>> have responded and for that I thank them.
>>
>> However on scanning through the remainder of the threads, it has become
>> fairly obvious that the vast majority of people here know a lot LESS
>> than I do.
>>
>> In particular a recurring theme is 'will I get better performance if I
>> change my ISP' and since the answer to this in most cases is a
>> resounding NO. I thought a dump of knowledge might be a good idea.
>
> Despite all your theoretical verbiage, the real answer is neither yes
> or no - it depends on where the speed restriction is. If the
> bottleneck is at the local exchange then changing ISP will have no
> effect unless, as you say, by switching ISP you also move to LLU.
>
> Practically, in my own case with nildram 'upto 8Mb' I had low average
> speeds (less than 1Mbps) during weekday office-hour for many months.
> Nildram told me it was my exchange and so no point in switching ISP.
> Despite that I decided to switch for a variety of reasons, e.g. because
> the new ISP was cheaper.
>
> The very same day I switched to an entanet reseller on the 'upto 8Mb'I
> went to average weekday office-hour speeds of over 5Mbps. In the 7
> months since then that good average speed has been maintained.
>
> As might be expected, the modem synch speed was the same with both
> ISPs.

that's the point Nildram had you on the high contention ratio backhaul.
Or your BRAS was too low..that can be reset tho.

>
> Somehow, at least in my own particular case, I find it hard to believe
> that the great increase in speed just coincidentally happened on the
> same day that I switched ISP. It is much more likely to be related to
> some ISP-related parameter (e.g. capacity).
>

Almost certainly it was either that the BRAS was set low..that can
happen if the line has been bad at some time, or Nildram didn't tell you
what contention ratio they had bought off BT.

It is possible to throttle speeds at the ATM level. BT will do that to
avoid flooding the DSLAMS, and the ISP may do that to provide a cheap
and nasty service maybe but its hard to see why..they will immediately
get found out. And the economics of doing it do not make a lot of
sense..thats not the big cost in the whole equation.


I am sorry I tried t be helpful. It seems that as usual fixed ideas and
egos are more prevalent than any desire to learn.


> Kit

tony h wrote:
> i agree with everyone (except OP)
> i changed from pipex to entanet and my internet experience changed
> dramatically, same sync speed but much faster actual speed day and night, no
> port throttling so i can use the internet as i like, low pings 24/7 which
> was NEVER the case before. clearly defined usage limits and penalties (with
> facility to check so far)., not just some vague FUP. freephone uk based tech
> support if i need it, not £1/min
> much as the technology to the isp is the same, from there on it's totally
> different.
> a simple analogy would be driving to ireland from uk, everyone would use the
> same road, but once at the pier choosing fast ferry or slow makes a major
> difference, both are boats, but one performs much better than the other.
>
>
Looks like it is as I said in the OP, that Pipex had you on a high
contention ratio backhaul..well that doesn't surprise me..with Pipex.

IIRC there are two contention ratios that BT offers. I had assumed -
obviously erroneously - that anyone on 2Mbps and up would be on the
faster one.

Ok I'll admit to being wrong on that point. I don't have an axe to
grind.Just trying to get the facts straight. And leern what I can and
pass on what I think I know.

> I am sorry I tried t be helpful. It seems that as usual fixed ideas and
> egos are more prevalent than any desire to learn.
>

sorry, i didnt know i had a fixed mind, or a large ego, i just reported the
experience that i had. i could have mentioned my neighbour, on a different
ISP (tiscali) who's connection slows to a crawl at certain times of the day,
while mine remains fast and usable (he still has 4 months to go on his
contract, then leaving), this has nothing to do with the exchange.

i'm always willing to learn, which is one of the reasons i subscribe to
this, and other groups, i'm also willing to have my convictions overturned
by anyone with a reasonable, verifiable argument, sadly, i have not seem one
on this thread yet, i have personal experience of a major improvement after
changing ISP, and of a neighbour having poor performance with a differnt
ISP, and i assume that the difference is the ISP, perhaps you could explain
what i'm getting wrong?

"The Natural Philosopher" <(E-Mail Removed)> wrote in message
news:(E-Mail Removed)...
> Kit wrote:
>> In article <(E-Mail Removed)>, The Natural
>> Philosopher <(E-Mail Removed)> wrote:
>>
>>> Good morning.
>>>
>>> I subscribed to this group yesterday because I was having a specific
>>> problem with my setup - see elsewhere - in the hope that here I might
>>> find some people with more knowledge than I had. Well so far one or two
>>> have responded and for that I thank them.
>>>
>>> However on scanning through the remainder of the threads, it has become
>>> fairly obvious that the vast majority of people here know a lot LESS
>>> than I do.
>>>
>>> In particular a recurring theme is 'will I get better performance if I
>>> change my ISP' and since the answer to this in most cases is a
>>> resounding NO. I thought a dump of knowledge might be a good idea.

[snip]

Just to add my 2p-worth ... it depends what you mean by performance:

In my experience, this means whether I get a prompt and intelligent response
to reporting a problem.

Two examples:

1) About 3 years ago a neighbour ordered broadband from BT. It took about 8
weeks from BT claiming that they had provided the service, to BT sending out
a technician who found a redundant filter on the line which when removed
actually allowed the ADSL service to function. Most of the intervening 8
weeks was spent in ringing the BT call centre (in India, of course) and
explaining the problem from the beginning every time. Mostly the answers
were of the form: "a fault has been found in the exchange, and will be
corrected in the coming week". There MAY of course have been faults in the
exchange, but somehow I doubt it !!!

2) Recently I complained to my ISP (Zen) that during most evenings the SNR
margin on my line dropped from 5dB to 2dB and while the router appeared to
be maintaining the connection, its throughput was negligible. This is with
a line attenuation of 62dB. I explained the problem, and within a couple of
minutes Zen agreed that they could ask BT to increase the SNR margin to 7dB,
saying that this might take a couple of days. In reality, within about an
hour, the router dropped the line and re-connected, achieving 7.0dB SNR
margin. Since then the SNR margin has varied a little, dropping sometimes
to 4.0dB, but the throughput has remained good.

This is why I would recommend Zen over BT; and I imagine many other
contributors have similar stories to tell.

So, in theory, it SHOULD not matter which ISP you choose, but in practise,
it DOES matter, if only for reasons of overall service quality.

--
Graham J

Graham J wrote:
> "The Natural Philosopher" <(E-Mail Removed)> wrote in message
> news:(E-Mail Removed)...
>> Kit wrote:
>>> In article <(E-Mail Removed)>, The Natural
>>> Philosopher <(E-Mail Removed)> wrote:
>>>
>>>> Good morning.
>>>>
>>>> I subscribed to this group yesterday because I was having a specific
>>>> problem with my setup - see elsewhere - in the hope that here I might
>>>> find some people with more knowledge than I had. Well so far one or two
>>>> have responded and for that I thank them.
>>>>
>>>> However on scanning through the remainder of the threads, it has become
>>>> fairly obvious that the vast majority of people here know a lot LESS
>>>> than I do.
>>>>
>>>> In particular a recurring theme is 'will I get better performance if I
>>>> change my ISP' and since the answer to this in most cases is a
>>>> resounding NO. I thought a dump of knowledge might be a good idea.
>
> [snip]
>
> Just to add my 2p-worth ... it depends what you mean by performance:
>
> In my experience, this means whether I get a prompt and intelligent response
> to reporting a problem.
>
I deliberately steered clear of two huge issues where ISPs differ WILDLY.

Server performance and support staff performance.

For me these are *the* reasons to choose an ISP - not raw speed.

But I didn't have time to say all that.