LinuxSA Mailing list archives

  From: Mark Newton <newton@atdot.dotat.org>
  To  : Jason Tan <jason@rebel.rebel.net.au>
  Date: Wed, 24 Jan 2001 11:59:11 +1030

Re: AOL? or best ISP

Before I enter this, I just want anyone reading it to take note of a couple
of things.

Most of the messages I've been sending in this thread haven't included
the name of my employer.  I'm not attempting to promote a particular ISP,
or a particular pricing plan, but I am attempting to stimulate a bit of
an evaluation of what it means to be a viable ISP in the real world.

I'm actually really surprised that there haven't been many contributions
to this thread by people from ISPs who aren't my employer --  I know
they're on this list, and I find their silence mystifying, since most of
the contents of this thread have been fairly general to all ISPs, not just
the particular one that I work for.  Please consider that when you're 
wondering why we've been talking about one particular provider for much 
of the past week.

Anyway, moving on:

On Wed, Jan 24, 2001 at 10:36:22AM +1030, Jason Tan wrote:

 > If they werent cheaper ISPs would not in most cases be caching.
 > Most dont do it for speed, they do it for profit.
 
I think that in once sense you're oversimplifying, and in another 
sense you're not recognizing that those two are quite similar.

Please let me illustrate with some ISP-neutral examples.

If any ISP has a cache which is delivering, say, 30% savings on bytes,
that means two things:

   - That's 30% worth of volume charges which they don't have to pay
     on volume charged links (if their links are volume-charged) -- this
     maps to your "doing it for profit" bit above, and it represents
     a 30% decrease in the "cost-per-megabyte" on average;  and

   - That's 30% worth of "free" capacity on a fixed-bandwidth link which
     they haven't needed to buy -- this maps on to your "doing it for 
     speed" bit above.

For example, a small ISP on a 2 Mbps connection will be able to behave
as if it's a 2.6 Mbps connection if they have a cache which is delivering
a 30% byte saving.  So in one way that's making things faster (they've
freed up 600 kbps), but in another way it's also making it cheaper (that
600 kbps isn't costing them as much).

Note that "not costing them as much" isn't the same as "not costing them
anything." 

Let's have a closer look at that small ISP.  Pretend they're connected
to Telstra with a Megalink 2Mbps circuit.  If you, as a walk-in consumer,
connect to Telstra you'll be able to buy bytes from them at 19c/Mbyte,
GST ex.  Now guess what their wholesale rate to ISPs which will compete
with them is?  You guessed it, 19c/Mbyte GST Ex!

So, there's this ISP.  It's getting a wholesale access deal which is 100%
identical to the retail access deal which you could get if you didn't use
them as your ISP, *and* that's before they factor in overheads, profit 
margin, etc.  So, in leiu of other factors, there is absolutely no reason
for a customer to use that ISP.

So the ISP installs a cache, which effectively cuts their wholesale price 
per megabyte to 14.6c/Mbyte (30% down).  They factor in overheads, profit,
infrastructure, etc, and come up with a retail figure of 18c/Mbyte (GST Ex),
but that's on a cut-thin-to-the-bone margin, and only if they sell A LOT
of megabytes:  It's only *JUST* profitable.

... then you come along and say, "But what about caching?  You should be 
selling it for 14c/Mbyte + GST!"

It just doesn't add up, Jason.  To do what you're doing would involve running
at a loss, at which point the ISP might as well not be there in the first
place.  When you're having that kind of discussion, remember that the 
wholesale rate is already higher than the retail rate, and the only reason
it can continue like that is due to caching, use of local peering points,
etc.

Note that the figures above have been made-up (i.e.: I'm not presenting 
you with a real-life case study), but a quick look around the marketplace
shows that they're pretty spot-on.  You'll find that variations in that
cost are due to ISPs getting varying wholesale prices by using providers
other than Telstra for their upstreams (including peering points), rather
than variations in cache efficiency.

Also note that the "good-fast-cheap, pick any two" principle hasn't been
applied here at all -- This mythical ISP isn't really making enough money
to have the luxury of making that decision on the profit margins used
here, they'll be lucky enough to just keep hanging on for dear life.
They're not making enough money to use the latest equipment and employ
good staff ("good"), they're not making enough money to plough it into
infrastructure ("fast"), and they're not big enough to have the economies
of scale you need to start dropping the wholesale price by playing one
wholesale provider off against another ("cheap").  Hopefully that helps 
you to understand why so many ISPs start up with a bang, then quietly 
whimper themselves to death over the course of the next year or two.

 > Alternately you could just charge a lower price for all bytes, based on
 > the average efficeny of your cache.
 
As I've shown above, this is pretty much what most ISPs do, and which is
why it's unsettlingly annoying to have people ask you to lower your prices
to take caching into account when the fact that you've already done that
is the only reason you're still in business.  How many iterations is that
process supposed to go through before the customer demands that the price
per megabyte drops to zero due to the effect of caching? :-)

 > Or do lots of oteh rhtings.
 > The real pointis that the internode guys keep telling about all the
 > expense, which soudns very reaosnable and is, but they are not mentioning
 > the techiniques they can and almost certianly do use to decrease actual
 > bytes downloaded.
 > Ie I dont think they are presenting a very accurate picture.
 
Excuse me?

You seem to have an impression that reducing bytes downloaded is the be-all
and end-all of a cheap Internet access plan.  That's simply not true, and
I think that if you're going to throw around accusations of inaccuracy 
you really should be justifying them.

For what it's worth, your usage of the plural "techniques" is misleading:
Apart from caching, there really *aren't* any techniques to descrease bytes
retrieved from upstreams.  And Napster, which constitutes a double-figure
percentage of traffic, isn't cacheable at all.  Factor that into the real-
world price of a $25 "leech account" and you'll see that it *REALLY* doesn't
add up.

 > Lots of peopel here woudl kow this, whether they woudl recall it or not
 > while reading the email, who knows.
 > I  just think that if a pricture is rpesented it shoudl represent the
 > whole picture, to soem extnet, and not just the bits that are convenient
 > for your argument.

What bits haven't been covered?

 > Yes but every time that someone mentions one of these ISPs that use a
 > dffrent combination to gpood and fast, that person gets a interndoe person
 > telling them why this is nto good, or how this is unsustainable, which is
 > blatantly not true in many cases.
 
I take it you're referring to me, since I've been the one using the word
"unsustainable."

I'm happy to debate this further (off-list if necessary), because you've
basically called me a liar above.  Where, exactly, have my accusations of
unsustainability been "blatantly not true"?  You must know, because you've
referred to "many cases" -- Right now I'm interested in knowing whether
these "many cases" are occasions you can actually cite, or whether you're
just blowing smoke.

As far as the historical record goes, I think there's ample evidence of
imploding ISPs to show that what I've been saying has quite a bit of 
merit.  But maybe you know something I don't.

 > And when someone does poke holes in or ask quesitons about  the
 > statements made, which lead to areas which may not be so favourable they
 > ignore it.

Again, you're going to have to cite some examples, off-list if necessary.
I've tried my best to give complete answers, while attempting to work 
within the graciousness which the mailing list has demonstrated by the
fact that this thread has been allowed to continue at all (and now that
Richard has suggested drawing discussion of my employer to a close I've
done just that -- my discussion here is generic to any ISP in Australia).

But I suspect the real reason we're having this conversation right now
is that you have a less-than-complete understanding of what it costs to
run an ISP.  A word of advice:  Instead of looking at a retail price and
working backwards to estimate what the likely wholesale price must be,
why not look at the (publically available) wholesale prices and determine
what you'd be retailing the service for if you were running it.

   - mark

--------------------------------------------------------------------
I tried an internal modem,                    newton@atdot.dotat.org
     but it hurt when I walked.                          Mark Newton
----- Voice: +61-4-1620-2223 ------------- Fax: +61-8-82231777 -----

-- 
LinuxSA WWW: http://www.linuxsa.org.au/  IRC: #linuxsa on irc.linux.org.au
To unsubscribe from the LinuxSA list:
  mail linuxsa-request@linuxsa.org.au with "unsubscribe" as the subject