I think it is a non-issue as well...  While there are always exceptions, we 
have been oversubscribing gig linked edge switches for years...  having 48+ 
people connected to switches that have a single uplink that is rarely or never 
tested.  And with the exception of some very dense places in wireless, you'll 
normally have less connected per AP.   For me, the focus needs to be on less 
contention on the airwaves, not throughput with new wireless.

Ryan H Turner
Senior Network Engineer
The University of North Carolina at Chapel Hill
CB 1150 Chapel Hill, NC 27599
+1 919 445 0113 Office
+1 919 274 7926 Mobile

-----Original Message-----
From: The EDUCAUSE Wireless Issues Constituent Group Listserv 
[mailto:[email protected]] On Behalf Of Frank Sweetser
Sent: Tuesday, March 24, 2015 11:06 AM
To: [email protected]
Subject: Re: [WIRELESS-LAN] 1GBE as a bottleneck to APs?

Personally, I'm not too worried about it.

While naively adding up the wireless marketing sheets gets you to > 1Gb 
numbers, especially when treated with Wave 2 pixie dust, I think there are a 
few factors which make this a low concern.

  - The wireless numbers are half duplex, while that 1Gb wired connection is 
full duplex.  This means that while your client bandwidth is probably going to 
be biased download more than upload, the upload and download packets that are 
bottlenecked through the common air time each have their own contention-free 
1Gb channel once they hit the wired network.

  - Wireless throughput is *very* picky at top speeds.  I've seen estimates 
that those magic wave 2 numbers won't be reachable more than a few meters away 
from the AP.

  - It only takes a few legacy clients hopping onto your nice new 11ac AP to 
drag you back down to a fraction of your peak throughput.  Given how many 
budget laptops are being sold today with 2 stream, 2.4GHz only 11n adapters, 
this problem will be with us for a long time.

Even if you do end up in a situation that legitimately needs over 1Gb, I'd be 
careful before relying on the LACP based solutions.  Unless you're terminating 
your user sessions locally, all of the traffic will be going through an 
encapsulated tunnel between the AP and controller, which can easily end up 
hashing all of the traffic down one link.  There are tricks to work around this 
(I believe Aruba opens up multiple tunnels with different endpoint IP 
addresses, for example), but this it's still an imperfect solution where 1 + 1 
!= 2.

So my guess is that we have a few years before it's a major concern, and I'm 
waiting on a decent answer for 2.5Gb switching before I do any real investment 
in a solution.

Frank Sweetser fs at wpi.edu    |  For every problem, there is a solution that
Manager of Network Operations   |  is simple, elegant, and wrong.
Worcester Polytechnic Institute |           - HL Mencken

On 3/24/2015 10:37 AM, Hinson, Matthew P wrote:
> I've seen a few articles here and there regarding possible solutions 
> for "the gigabit bottleneck" as it pertains to .11ac access points. 
> Said solutions include Cisco's forthcoming protocols for 2.5G and 5G 
> over CAT5 cabling as well as LACP'ing two gigabit ports per switch and AP as 
> some vendors suggest...
>
> My question for the group is: Has anyone actually seen a throughput 
> issue using gigabit to the edge? Certainly your distribution layer 
> gear could be a limitation if it's not specced correctly, but I've 
> just never seen a situation where I've wished for more than 1000BASE-T 
> to an AP. Our fastest 802.11ac access points can "only" hit 
> 600-700mbit/s real TCP throughput, and that's in ideal, almost laboratory 
> conditions.
>
> Thoughts?
>
> Thank you!
>
> Matthew Hinson
>
> Network Operations
>
> ********** Participation and subscription information for this 
> EDUCAUSE Constituent Group discussion list can be found at 
> http://www.educause.edu/groups/.
>

**********
Participation and subscription information for this EDUCAUSE Constituent Group 
discussion list can be found at http://www.educause.edu/groups/.

**********
Participation and subscription information for this EDUCAUSE Constituent Group 
discussion list can be found at http://www.educause.edu/groups/.

Reply via email to