off-topic: historical query concerning the Internet bubble
kaze0010 at umn.edu
Fri Aug 13 23:01:20 UTC 2010
Roland Perry wrote:
> Kenny Sallee writes
>> So the whole 'myth' of Internet doubling every 100 days to me is
>> something someone (ODell it seems) made up to appease someone higher
>> in the chain or a government committee that really doesn't get it.
> [Whether it was really 100 days, or 200 days...] a statistic like this
> has very real operational significance, because it sets expectations
> in the minds of senior management and investors that the new shiny
> hardware (or leased line, or peering agreement...) you just put in
> place isn't going to last "a lifetime", and will need
> replacing/upgrading really quite soon.
Part of this rapid hardware replacement cycle almost certainly had to do
with the rapid growth in CPU capabilities in comparison to software.
New classes of applications and capabilities were opening up just as
fast as CPUs would allow. Many network appliances use embedded
processors based on the same chips used in desktops or laptops of
similar vintage. They run custom software and may have additional
dedicated chips. Thus the development of the FrankenPix and custom
Linksys wireless router firmwares.
Today even a 3(+) year old machine can do a fine job running office
tasks (given enough RAM), whereas in the late 90s/early 00s, a three
year old PC was not likely to be able to run the then current software
very well. Today CPUs have progressed so far ahead of most software
that we're able to combine multiple systems into one through
virtualization and still obtain good performance. Tasks formerly given
to dedicated chips (RAID, sampling rate conversions, compression) are
now commonly done on CPUs and GPUs.
I also recall articles/webpages/blog-precursors talking about how many
packets a particular CPU could route per second. The articles might
have been in relation to building custom Linux based routers and router
hybrids (such as router-bridges, adding QoS, etc.)
I feel that recently many changes in information technology have become
less revolutionary and more evolutionary as we look for the reasons to
build newer/faster/stronger/better equipment. The rise of netbooks as
low CPU/GPU power machines underlines the evolutionary changes. The
next series of revolutionary changes are still waiting in the wings
(compact/portable devices, realtime 3D, gaming, scientific, and
rendering applications are still pushing the envelope).
> Another meme at the time (at least in the UK) was the idea of
> "Internet Time", where things happened four times as fast as "real
> life". So you'd realise that things like a "five year plan" were
> really only going to last just over a year. And, of course, policy and
> law related to the Internet gets out of date four times as fast, too.
I know organizations where equipment refresh/purchase cycles have been
stretched from 3 years in the early 2000s to 5 years now, as they've
observed both a slowing in need for the latest and greatest, as well as
this being a response to budget pressures. Replacement periods are
becoming less based on technological obsolescence than on equipment
failures and end of warranties.
More information about the NANOG