Internet II is coming...

Jim Browning jfbb at atmnet.net
Thu Oct 10 19:48:33 UTC 1996


>From:  Robert E. Seastrom[SMTP:rs at bifrost.seastrom.com]
>Sent:  Thursday, October 10, 1996 6:21 AM
>
>2) Copper pairs to the telco (both in the POTS scenario and the
>D4-framed HDSL DS1 with a channel bank scenario) suit my purposes for
>switched voice just fine.  The concept of getting phone service from
>an organization like a cable company or an ISP kinda leaves me cold --
>I want my phones to be more reliabe than that.  What is your motive in
>proposing to fix something that ain't broke?

Consolidating a customer's investment in local loop circuits enables the 
purchase of a larger pipe ($$ leverage), and enables that circuit to be 
used more effectively by taking advantage of the fact that each service is 
not used 100% of the time.  The effect of this is to encourage broader use 
of higher-capacity local loops, thereby cranking up the average speed of 
'last mile' connections to the Internet.  This makes the introduction of 
new applications more viable, which in turn helps to feed the continued 
growth of the Internet, which I consider A Good Thing.

But I suspect you've heard this before...   :-)

>3) If you try to make a vehicle that will both haul a few tons of logs
>and handle like a Lotus in the turns, you will end up with something
>that does neither well.  This is what you have when you try to run
>voice and data across *any* network, not just ATM, but IP as well.

And given enough resources, one can choose to build separate networks for 
each type of service.  I believe MCI has done just this, although their 
data network now does contain ATM.  However, At least one large carrier, 
just announced plans to migrate both their voice _and_ data networks to 
ATM, including support of LANE.

Go figure...

>4) The most often touted advantage of voice over ATM is bandwidth
>savings, not having to have dedicated bandwidth per circuit, etc.  But
>take a look at the decay curves of (a) the cost of bandwidth, and (b)
>the cost of ATM switches, and tell me how much longer it's gonna make
>any sense to shell out all that extra dough.

These are important factors, however it will always be an economic 
advantage to effectively utilize purchased bandwidth, regardless of how 
much is available at what cost.  What will happen is that integrated 
services will become affordable to smaller and smaller businesses.

>5) Decay in the cost of bandwidth notwithstanding, in an Internet
>network given the current explosive growth, even a 15% cell tax is
>unacceptable.  Not to mention that native IP networks have a much more
>graceful performance degradation curve when they get congested than IP
>over ATM networks.  Not to mention, saturation comes 17% later.

However, many operators have realized that the best way to avoid congestion 
is to get the fattest pipe they can, and that remains an ATM pipe.  If 
bandwidth becomes as cheap as is necessary to support your arguments, then 
you can stay ahead of congestion with _either_ approach.  I prefer to do 
this as while maintaining the ability to provide everything IP has to 
offer, with the added benefits (raw speed, voice) that ATM brings to the 
table.

>Well, I really have to run, but I don't think the above comments boil
>down to "ATM sucks".

I agree.  This is now just good old fashioned argument...
--
Jim








More information about the NANOG mailing list