Re: Edge Bandwidth: 18mbps connectivity widely available in U.S., we just don't realize it.

["Dave Farber" <dave () farber net>]

From: Bob Frankston [mailto:bob37-2 () bobf frankston com] 
Sent: Sunday, November 25, 2007 11:21 AM
To: dave () farber net; ip () v2 listbox com
Cc: 'Thomas Leavitt'; 'Lauren Weinstein'
Subject: RE: [IP] Edge Bandwidth: 18mbps connectivity widely available in
U.S., we just don't realize it. 

 

I'm happy to see a new generation of dual WAN NAT/Routers appearing. The
last batch was in 2004 and I used my Xincom for years. I recently got a
Draytek but it was missing a key feature I need - local DNS override because
local addresses are different from global addresses. I won't go into detail
on this issue except to note that this is one reason we need protocols that
are name<=>name rather than address<=>address. 

 

I'm very interested in other devices - last time I looked there were few new
ones after the 2004 dual WAN bubble collapsed (AKA, there didn't seem to be
a huge market).

 

If you have multiple lines from a single carrier bonding can work to get a
higher effective speed but that should be a technology that can be deployed
transparently just like other ways the network operator provisions the
network. My real interest is in bonding at the edge - taking a Comcast and
Verizon connection and using both for reliability and higher aggregate
speed. And to work around port blocking by using Comcast for inbound when
Verizon frustrates connections.

 

The other big problem with today's protocols is that you can't get the
user-bonded speed for a single TCP connection - it's only for the aggregate
since each connection has to stay with a single pipe (unless you use
something like Bit Torrent). This is another reason I want the name<=>name
protocols.

 

As an FYI, despite my reservations about FiOS Verizon does offer 50Mbps down
so you really want routers capable of 100Mbps or more - especially outside
the US. The other caveat, however, is that high speed on the local link is
very different from high speed to another point on the network. As I point
out in http://www.frankston.com/?Name=InternetDynamic the real speeds are
far lower. On the other hand using multiple connections and/or Bit Torrent
from MIT which is peered locally I did achieve the full speed.

 

I notice that the Peplink is limited to 25Mbps for the consumer version but
for a much higher price you can get faster speeds for expensive DSL lines.
It highlights the disparity between the 50Mbps I can get from Verizon and
the limited speeds available elsewhere. We tend to judge reality by our
local experience so I can blithely assume 50Mbps is available and, at worst,
I can got 16Mbps from Comcast or RCN. Others are struggling to get 1Mbps.
Again, that's a topic in its own right. As Thomas' post shows by getting
control at the edge we start to shift the dynamic. Imagine if we owned our
local infrastructure .

 

I cc'ed Lauren because this is an NN issue - who determines the capability
of the network and what happens as we seize effective control at the edge.
For those doing measurements it adds some more complexity as we have the
policies of multiple carriers beating against each other as each thinks it
understands the traffic based on the presumption that it's predictable like
telephony traffic used to be.

 

 

-----Original Message-----
From: David Farber [mailto:dave () farber net] 
Sent: Sunday, November 25, 2007 09:11
To: ip () v2 listbox com
Subject: [IP] Edge Bandwidth: 18mbps connectivity widely available in U.S.,
we just don't realize it. 

 

 

 

Begin forwarded message:

 

From: Thomas Leavitt <thomas () thomasleavitt org>

Date: November 25, 2007 5:36:15 AM EST

To: dave () farber net

Subject: Edge Bandwidth: 18mbps connectivity widely available in U.S.,  

we just don't realize it.

 

Dave,

 

I've been noticing that prosumer level router/firewall combo devices  

(under $500) have started integrating dual WAN interfaces (with  

bonding capability) as a standard option. In the process of evaluating  

solutions for providing a consulting client of mine with options for  

expanding their WAN capacity, it occurred to me to wonder if someone  

had taken this to the logical next step, and created an affordable  

device that bonded more than two WAN interfaces. Lo and behold, a  

brief search in Google produces this product: the PePLink Balance 30,  

which load balances *three* WAN interfaces, and costs less than $500.

 

http://www.peplink.com/products/balance-30/

 

There's a business class version with more features, the 300, that  

costs $995.

 

I'm sure there are other devices out there. This means that your  

average prosumer, small office is no longer limited to a choice  

between 6.0 DSL and 10 mbps shared media cable internet. For under  

$200/mo., you can get an 18 mbps connection, anywhere DSL is generally  

available. It also means that, if ADSL2 is available in your area, you  

can get up to 45 mbps download speeds, for under $400/mo. (c.f.  

launchpad.net's 15/1 ADSL2 service @ $129.95/mo.,
http://www.launchnet.com/adsl2+_dedicated_soho.php 

  ). Upload speeds are still pathetic (1.0 per bonded line), but  

considering that people still pay $359/mo. for a T-1 line, $390/mo.  

for twice the upstream connectivity starts looking pretty good.

 

Gotta wonder why no ISP has bought these things at a discount, in  

volume, and started selling "18.0 DSL" (or "45/3 DSL") as a standard  

service. Is there some technical issue that makes installing three DSL  

lines to a typical premises difficult? No indications to that effect  

in my experience.

 

Regards,

Thomas Leavitt

 

 

-------------------------------------------

Archives: http://v2.listbox.com/member/archive/247/=now

RSS Feed: http://v2.listbox.com/member/archive/rss/247/

Powered by Listbox: http://www.listbox.com

-------------------------------------------
Archives: http://v2.listbox.com/member/archive/247/=now
RSS Feed: http://v2.listbox.com/member/archive/rss/247/
Powered by Listbox: http://www.listbox.com