[*] [Fwd: [CAnet - news] Tool for measuring network neutrality]

[Ron Dallmeier]

The following was posted on the "CA*net News" list. I thought it might
interest members of this list so I am cross-posting it.



http://blog.tomevslin.com/2006/11/hope_for_intern.html

Hope for Internet Quality (and Neutrality) Measurement

Fellow My Way blogger Eric Hernaez recently posted the important news that
Cisco and Brix are proposing an extension to the SIP standard which allows
devices to be put in a loopback mode in order to measure the quality of
their connection to an Internet service.  Standards are usually pretty
dull stuff but this proposal could be a very big deal - even if the IETF,
to which it was submitted, takes its usual forever to adopt it.

Although Cisco and Brix propose the extension in order to give VoIP, real
time text, and video over IP providers a way to diagnose customer
problems, the potential is much greater than just these very worthwhile
applications. This technology, if widely deployed, can be the answer to
two vexing problems: 1) how do consumers cut through the hype to compare
the quality of broadband providers; and 2) how do we know if ISPs are
violating the principals of network neutrality (these principals are not
enshrined in any law or regulation), perhaps to advantage their own
applications over those provided by competitors.

As Eric points out:

"The Cisco collaboration is a major coup for Brix. With the industry
leader adopting this method of link testing, it will become a de facto
standard whatever happens in the IETF. I would be surprised if other ATA
vendors [nb. besides Cisco] did not quickly fall in line. Look for
softclients and web-based tools to implement the draft in the near
future."

The devices which implement SIP are in almost every home or office which
uses VoIP.  SIP is built into routers which have VoIP capability and into
cheap little VoIP boxes called ATAs (what connects your phone to the
Internet if you have a service like Vonage).  SIP devices are literally
everywhere and spreading.  SIP is also implemented as software in many
IP-based VoIP applications.  A new generation of mobile phones uses SIP
when connected to a WiFi hotspot.

An article in last Saturday's NY Times says:

"Determining the speeds consumers are actually getting is tough to
measure. Cable speeds can vary if many people in one neighborhood are
online at the same time, like after dinner. Access over phone lines can be
slower if the customer is far from the switching office, where the
Internet signal originates."

The article also says:

"In many cases, consumer advocates and industry analysts said, customers
do not get the maximum promised speed, or anywhere near it, from their
cable and digital subscriber line connections. Instead, the phrase 'up to'
refers to speeds attainable under ideal conditions, like when a D.S.L.
user is near the phone company's central switching office."

Both of these assertions are true.  There are many other factors which
also affect the bandwidth (aka speed), latency, and jitter experienced by
a particular user at a particular place.  Up until now (despite the
valiant efforts of some online connection test sites) it's been
impractical to gather enough comparable data to assign grades to the
various providers of broadband service.

OK.  Fast forward.  Now we've got millions of devices distributed
throughout the country and the world which can cooperate in automated
measurement of bandwidth, latency, and jitter between themselves and some
host service somewhere. Let's assume that the owners of these devices (you
and me and all our friends) consent to have a trusted organization put the
devices in feedback mode every once in a while for a few seconds for the
purpose of gathering some data.

The cure for lots of variables is lots of data points.  What is almost
impossible to judge anecdotally becomes very tractable to measurement if
it's measured enough times in enough places.  "Individual performance may
vary" but it will become very easy to know which broadband providers
perform how well where and doing what.  It'll be much easier to know if
you want to buy your broadband access from the cableco, telco, or the new
WISP (wireless ISP) down the street.  When someone advertises "speeds up
to..", it'll be possible to know how often that speed or 95% of it is
actually obtained. With enough data we can distinguish between websites
which are slow to respond and networks which are slow to deliver the
response.  Most packets traverse multiple networks; millions of
triangulations on packet delivery performed by millions of devices should
show which networks slow packets down or lose them and which speed them on
their way.

I proposed a citizen journalism project which started with the development
of software which could be run by millions of volunteers to determine both
network performance and possible violations of Internet Neutrality.  Maybe
software developed to this new SIP extension fills the bill.  I'll be
interested to learn more.

If it does do all I hope and if this capability ends up standard in
routers and ATAs, it solves at least part of the development problem and
much of the distribution problem for the citizen journalism exercise.  For
technical reasons, it is much better to run the monitoring software on a
router connected directly to the broadband provider's modem than on a PC -
especially if that PC has a WiFi connection.

SIP updates can often be accomplished by a firmware download so it's
conceivable that this capability can be added to existing routers and ATAs
in a routine maintenance release or upgrade (note:  I haven't done the
technical work to know whether there may be some obstacle to this).  It's
also possible that SIP capability will be added to non-VoIP routers and
devices specifically to enable this loopback capability.

We may be able to compare networks AND look for non-neutral network
connections sooner than I thought.