[AFRINIC-rpd] IP Address / Wireless concurrency

[Andrew Alston]

Hi Guys,

Over the last few days one of the questions I have had to deal with when
working with IP address applications surrounds concurrency of usage of
networks.  This is particularly relevant in the case of wireless networks,
where space is being applied for and the infrastructure is not yet in
place.  I.E A particular institution has ordered a wireless network of a
particular size, to serve a community of X number of users, from provider
Y.  The institution does not use NAT for various reasons that are out of
scope of this discussion.

Now, this raises some interesting questions, when motivating for IP space
for this wireless network, how do you calculate potential IP concurrency
for the network, and what guidance can we give to AfriNIC around the
concurrency issue to help them as they process our applications.  It's a
curious question, but its become very clear to me in recent discussions
that this is something that needs to be addressed.

So, let me state what I have found in research I have done (I physically
visited 3 universities in the last 2 days and spent a lot of time with them
and looking at numbers they had).  I openly state, in the cases of these
universities, they were urban area institutions we were looking at, student
base I would say would probably fall into the "middle class" category.
 (This is relevant considering other research which I will expand on).

We found, in such institutions, particularly where individuals were
required to authenticate to get access to the network, we were running at
an average concurrency per individual on site of 1.9:1 addresses.  Looking
more closely, this was due to the smartphone prevalence (with at least 50%
of leases in every case being smartphones), and the rest being a balance of
notebooks/ipads/android tablets etc.

The concurrency of leases dropped significantly when looking at data
surrounding more rural institutions, however, the smartphone prevalence
remained, and the ratio was consistently over 1:1 per people in range of a
wireless AP.

So, before I write further, my question to all of you that are running
wireless networks, what sort of address concurrency are you seeing, and
what are the stats you can provide so that we can look at some sort of
uniform way in which to address this question in the applications process.

Of course, that also raises the question (particularly relevant on a
university campus), about concurrent numbers of people on campus.  In an
institution with 45 thousand full time students, what percentage of those
people are on campus at any one time.  Currently, in initial (and albeit
very brief) research, we're seeing this at around 80% at peak (and on
wireless to avoid running out of leases, peak is what you are catering
for).  I'll send some more research data back on this once and if we manage
to gain firm figures (these figures will be gained by looking at
universities who run access control onto campus and we can look at the
access control to get figures).  Again, I'd be interesting in hearing from
others here involved in these types of networks what stats they have.

As a matter of interest, by our calculations, on a university campus with
50 thousand people (5 thousand staff, 45 thousand students), with 100%
wireless coverage, we hit an ip address assignment concurrency calculation
of 76 thousand addresses on the wireless with current data that is still
being expanded.

For the purposes of this discussion and to avoid going off topic, please,
let us avoid the debates about NAT on wireless networks and other such
things, I am merely looking at the discussion of IP address concurrency.

Thanks

Andrew
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.afrinic.net/pipermail/rpd/attachments/20121009/7eff3449/attachment.html>