Many school
districts want 802.11ac networks for their schools. Some are even willing to rip out there still
very useful 802.11n networks for the chance to get the great speed increase
that 802.11ac provides.
What no one told
these districts is the great speed gains they were promised is all marketing
and fluff. Ok so it's not all marketing and fluff but the configurations needed
to accomplish these data rates is not practical for a high density school
deployment.
School districts in
my area have been going crazy lately with standardized online testing and video
streaming which has led to the explosion of devices in the classrooms and on
wireless networks. Some schools have even
deployed a Chromebook to every student.
Vendors often come
into schools looking at 802.11ac and sell them on the big data claims of AC. We
have all read about 80MHz and 160MHz channels with "X" number of
spatial streams and 256 QAM netting us well over a gig of data. Not to mention
MU-MIMO which I kept hearing would be "switch like". With data claims over one gig districts are
thrilled to hear that the wireless clients would have higher throughput then
wired and they were willing to pay anything to get it. Once vendors had the
districts hooked on AC data rates they then turned around and told them what it
would take to get there. For most it
wasn’t just an AP overhaul which sometimes included a new controller or two.
With the need to have a multi gigabit backhaul this meant that districts were
going to need to run an additional drop or in some cases 2 new drops to every
location an AP was going to be deployed. Which than meant switch closets would
need additional port density… yup new switches!
So for all this what
did the districts get… the network they asked for. I wouldn’t say that these
networks performed poorly (some did) but I would say it was overkill. So why is
this overkill? If you take a look at the networks as a whole, things start to
fall into place. Most districts I work with have moved to cloud based
applications and some have even gone as far as to get rid of all the file
servers in the district (they moved the servers to our NOC with a 1 gig link
back to the district). Between switch closets most have 1 gig links to the core
but are slowly migrating to 10 gig. All this makes the internet or WAN link the
critical path. What's the point of
increasing the wireless data rate and AP backhaul if you're just going to
bottleneck it somewhere else? Most of the districts I work with have nowhere
close to 1 gig of internet. Then we need to take a look at the clients they
have deployed. Lots of schools are moving
away from expensive laptops and deploying Chromebooks instead. Why? Because
they are cheaper and everything they do is in the cloud anyway so why not! All
the Chromebooks that I have come across only support a max of 2 spatial
streams, MCS 9 and 80MHz which nets you 866mbps in 5GHz. Sounds like we are
getting close to needing that 2gig backhaul now.. Add in a client (Chromebook)
on 2.4GHz transmitting at the same time with a 40MHz channel (this is a no no)
and an MCS of 15 for its 2 spatial streams and you get 300mbps
(866+300=1166mbps). Finally we break the 1 gig backhaul and our network is
justified! Or is it? Don't forget to take into account that the data rate is
not the throughput. Wireless has a lot of overhead that never makes it to the
wire so let's assume 70% throughput on the wire (70% of 1166 = 816) and
suddenly we are back down below 1 gig.
Now given the nature
of wireless let's look at a more likely scenario in our high density
deployments. All APs would need to use
20MHz channels in both 2.4 and 5GHz do to the need for more contention domains.
A Chromebook with 802.11ac using 2 spatial streams and an MCS of 9 in 5GHz and
another using an MCS of 15 in 2.4GHz transmitting at the same time would give
you an over the air combined data rate of 317mbps. At this point I don’t even
need to figure the wired side throughput to determine that we don’t need more
than a 1 gig backhaul. On top of all
this we didn’t even get into the application throughput requirements which
would be well below our over the air data rates.
No comments:
Post a Comment