As the go-to source for all news
relating to bufferbloat, I’m glad to announce that the
first of several possible solutions to the problem
will shortly be available, just in time to save the
Internet from self-destruction.
What, you didn’t know the Internet was
self-destructing? Well it is.
Bufferbloat, my #1 prediction from 2011, is an
artifact of cheap memory and bad planning in the
Internet Age. In order to keep our porn streaming
without interruption we add large memory buffers in
applications, network cards or chipsets, routers, more
routers, and even more routers until the basic flow
control techniques of the TCP protocol are completely
overwhelmed. Data glugs through the system like a gas
can with no vent. Our solution to date has been to
make our pipes (and therefore our glugs) bigger, but
in the long run that won’t help. Latency increases and
performance declines.
Many Internet users are unaware of bufferbloat
because it has been masked by faster computers and
bigger pipes and because it sneaked up on us slowly
over time. But here’s a test. Think back to your
first broadband cable or DSL Internet connection,
right after you finally got rid of dial-up. How much
faster is your Internet connection today than it was
back then? Don’t think in terms of numbers but of
subjective performance. It’s not much faster at all,
is it? That’s bufferbloat.
My first broadband connection was a Northpoint DSL
line back in 1996 running at 384 kilobits-per-second
and my primary computer at the time was running at
something like 40 MHz. My Internet connection today
is 24 megabits-per-second (about 70 times as fast) and
my main PC has quad cores running at three GHz (500
times as fast!). My Internet connection should
certainly feel at least 70 times faster. Yet for most
purposes my connection doesn’t seem any quicker than
it did back in 1996. Part of this is that we are
pushing more bits through the pipe, but the rest is
bufferbloat.
The solution to bufferbloat is Active Queue
Management (AQM), which is controlling buffers to
maximize data throughput. The most effective thing
most of us can do right now to reduce bufferbloat is
to manually set all our buffers at the smallest
possible size, but that’s just a coping strategy, not
AQM. AQM will dynamically adjust buffers to improve
network efficiency with the result that our Internet
connections will all speed up as if by magic.
But as it was with Tinkerbell, such network health
can happen only if we all believe. To
eliminate bufferbloat it isn’t enough for you and me
to change our ways, we all have to do it.
And we finally are getting the tools to do so. This
morning the first public bufferbloat solution appeared
in a paper on an ACM website titled Controlling
Queue Delay by Kathleen Nichols and Van
Jacobson. If you read the paper please also look at
today’s blog post from bufferbloat
pioneer Jim Gettys explaining what it’s all about.
But the best is yet to come because within days or
weeks we’re likely to see AQM-equipped beta code for
Open Source routers from many manufacturers. Depending
on how well this code works and how quickly
manufacturers like Cisco, Netgear and others adapt it
for their non-Open Source network hardware, this could
be the beginning of the end for bufferbloat.
It’s nice to print good news for a change.
Of course I sometimes get carried away, so here’s a
cautionary note from Jim Gettys: “We have only
ethernet running. Wireless will be *much* harder,
and take months to years to get working properly.
This has to do both with the much more complex
queuing needed for AP’s, and the hair that packet
aggregation causes (802.11n). Also CeroWrt only runs
on a small amount of hardware at the moment;
and CoDel only on its ethernet so far. Also lots of
other things besides home routers need fixing: e.g.
your cable/dsl modems, FIOS gear, etc…”
Like he said.
My advice to Cisco, Netgear, D-Link and others is
that this could be an important moment in their
businesses if they choose to approach it correctly.
It’s a chance to get all of us to buy new routers,
perhaps new everything. Think of the music
industry bonanza when we all shifted our record
libraries from vinyl to CDs. It could be the same for
networking equipment. But for that to happen the
vendors have to finally acknowledge bufferbloat and
use their marketing dollars to teach us all why we
should upgrade ASAP. Everybody would win.
Take our money, please.