Re: [Nethistory] Measuring the Internet: When History Means Measurement

[Curtis Villamizar <[e-mail]>]

Good luck with that.

ISPs have always been notoriously somewhat secretive about their
network topology, even more secretive about the capacity of their
network, and extremely secretive about the actual traffic loads.

Attempts to document all Internet outages (you wrote shutdowns - wrong
word) over time may prove even more difficult.  Small outages are a
constant occurance and not even worth noting.  Routing errors can be
vexing for end customers because they are persistent and can be missed
by ISPs (and can be caused by ISP misconfiguration).  Any such outage
is often attributed to just the organization that reported the outage
or identified the problem even though the impact might have been
widespread.  These days there are also IPv4 vs IPv6 outages (an outage
exists for one but not the other) often due to a routing error.

There was good information collected on the NSFNET when that was a
thing and somewhat central to most of the Internet, but that is now
ancient history (and not a complete look at the Internet even back
then).  Check with Merit for archives and the IMR.  Also check CAIDA
to see what they have.  You may find useful information on routes
going away from the route view servers and possibly archives.  Some of
the regional NOGs or NICs produce route churn graphs and may have
archives.  There are many sources of information and many
methodologies for collecting and analyzing it.

As the number of ISPs grows the difficulty in determining level of
network traffic becomes increasingly intractable.  When routes go away
(or harder to detect blackhole or loop) it is hard to know how much
traffic is affected.  Not all prefix outages are equally important and
the number of addresses within the prefix is a poor estimator (almost
useless for IPv6 with many ISPs freely handing out /48s).

Whenever you talk about outages, you need to quantify the extent of
the outage.  Small outages are a constant.  You have to decide what is
a big enough outage to count as significant and then even harder you
have to decide on a way to measure extent of outages.

Curtis


In message <[e-mail]>
Guntur Wiseno Putra writes:
> 
> Dear nethistory,
> public-webhistory &
> public-informationarchitecture
>  
>  
> * Measuring the Internet: When History Means Measurement
>  
> It is about a "reliable place to find answers to questions about the
> current state of the Internet... (but at the same time it is about
> attempting historical values as it is also about a)... single site or tool
> that brings together the multiple measures required to understand the
> Internet=E2=80=99s evolution and health.
>  
> ..... A Web-based dashboard will present these trends and contextualize
> indicators within an overall narrative of Internet evolution and promotion
> of the Internet way of networking...".
>  
>  
> It is part of the 2020 Internet Society Action Plan Projects
>  
> https://www.internetsociety.org/issues/measurement/
>  
>  
> Regard,
> Guntur Wiseno Putra