[AusNOG] IPv6: "Objections to sale"
reuben-ausnog at reub.net
Wed Mar 6 20:51:41 EST 2013
On 6/03/2013 7:51 PM, Mark Newton wrote:
> Have any of you applied enough introspection to figure out why this particular
> piece of internet technology, among all others, is the one that you don't seen
> to want to touch?
So here's my geek story. Perhaps this may encourage others to have a
bit of a go too.
Before I ventured down the path of IPv6 the "excuses" I had were:
- The numbers look too big, I couldn't remember them off by heart
- I had no idea how subnetting worked in IPv6
- I had a mixture of equipment that could and couldn't do IPv6
- There was no real advantage in learning it as no one else was asking
for the information
I eventually decided to take the plunge, and started by moving my home
network over to dual-stack 3 or so years ago. It was relatively
painless - and having the key building block of an Internode service I
was able to properly test it out.
It was relatively easy. So easy in that that I had largely completed it
in a weekend. That was a mixture of Cisco devices (including Small
Business switches which unlike the SB phones, do work with IPv6), Linux,
Windows - a good spread of things. The notable exception being my Cisco
SPA525G2, but almost everything else was capable.
So, encouraged by that success I took the idea to work (at the time I
worked for an ISP as the Network Admin) and started to plan and roll it
out, having learnt somewhat from the lessons and confidence gained from
doing this at home. We had an existing AS and IPv4 allocation from
APNIC (so an IPv6 allocation was free and didn't require justification),
were able to get IPv6 PIPE peering, and were able to also get
international IPv6 transit via NTT at no extra charge. Nextgen
eventually followed suit and offered IPv6 transit too. Cost so far?
$0; just a bit of time.
I completely overlayed IPv6 subnetting atop of IPv4 subnetting, mirrored
our iBGP topology and external routing. So implementing IPv6 was really
just "adding" capability to the network, not redesigning anything. It
had one quite unexpected side effect - I was able to completely
implement IPv6 while connecting to everything over IPv4, and one time at
a later date when there was a mishap with IPv4 I was able to get to
everything I needed over IPv6. Bonus!
Now in hindsight the reasons I had to not look at IPv6 earlier were due
to misinformation and bad perception:
- I didn't actually need to remember the addresses, I started to use DNS
extensively, rather than mostly (which arguably I should have been doing
for IPv4 all along anyway as using DNS is a key recommendation in making
for an easy transition in either direction)
- Subnetting and routing was the same in IPv6 as IPv4, especially if you
maintain the same layer 2 topology - which I *highly* recommend
- A small number of devices which couldn't/wouldn't do IPv6 are left on IPv4
- It provided the marketing guys with something they could take to
customers and say "We do IPv6 and we have in house expertise to help you
with it", so internally it was a relatively easy sell
All up? Extra cost in terms of hardware: $0, extra cost in terms of
software: $0, extra cost in terms of monthly recurring services from
third parties: $0. No outage or impact to end customers (so no lost
goodwill), the only indirect cost was my time spent, as a full time
Without doubt the engineers I worked with then were of the view that
there was no real need to learn IPv6, but they sure knew how to disable
it on Windows. Most engineers would rather spend their time on
something which is more pressing and urgent and that they can use every
day. They would certainly rather learn something else that they can
take to the boss to show what they're doing to help the business become
more efficient, and IPv6 doesn't fit in in that (yet).
There may be a technical shortage of IPv4 addresses, but while
businesses in AU/NZ can still get IPv4 addresses easily, even with
justification, they don't want it and don't see the point of investing
time on IPv6, especially if that time is chargeable time obtained from a
third party integrator. It won't be until the boss is told his business
cannot make the internal web server visible to roaming sales reps, or
something similar that the penny will drop and we'll start to see more
interest in an alternative. Until then the IPv4 address shortage will
continue to be perceived as too far away and "somebody elses problem".
More information about the AusNOG