All Things 802.11ac - Question 6: What Needs to Be Done on the Wired Side of the Network?

  • 1
  • Question
  • Updated 5 years ago
  • Answered
No matter what technology is applied, increasing numbers of users and devices result in Wi-Fi placing ever-greater loads on the rest of the network’s infrastructure, opening the possibility of bottlenecks that defeat the purpose of 802.11ac. What are the key items on the wired side that should be on a network manager’s checklist as .11ac is rolled out? Switch capacity? PoE? Will greater than 1 Gbps be required for AP links, and is 10 Gbps enough for uplinks? What about network management – any special considerations there?
Photo of Craig Mathias

Craig Mathias

  • 63 Posts
  • 0 Reply Likes

Posted 5 years ago

  • 1
Photo of Nick Lowe

Nick Lowe, Official Rep

  • 2491 Posts
  • 451 Reply Likes
I think this is a really interesting, poignant question for high density deployments.
A 10 Gb/s distribution layer is often prohibitively expensive at the moment for many organisations which means a high upfront CAPEX. The warranty on such kit then typically does not include a NBD limited lifetime warranty seen on lower end equipment which raises OPEX too.

I really hope that things like Broadcom's Trident II start to bring the price down here as it gets introduced in to the market to make it all more viable.

I wonder if this is an opportunity for the wireless vendors who have introduced switching to steal a march on the more established, traditional incumbent players.

Often organisations also have issues going to a 10Gb/s uplink because of the use of OM1 fibre which requires further consideration.
Photo of Matthew Gast

Matthew Gast

  • 284 Posts
  • 63 Reply Likes
If you managed to survive the 11n transition without upgrading to 1 Gbps edge switches, getting that rectified is the first order of business. 10 Gbps uplinks are a practical requirement, and are readily available on any gig edge switch.
Photo of Nick Lowe

Nick Lowe, Official Rep

  • 2491 Posts
  • 451 Reply Likes
It's when you need to aggregate 10 Gb/s from the uplink ports on your access layer that it becomes more expensive though. A 10 Gb/s core isn't cheap either.
Photo of Matthew Gast

Matthew Gast

  • 284 Posts
  • 63 Reply Likes
Much of what's been written about planning -- even by me -- says that you should start by understanding what's on your network. Fortunately, a whole new series of application level visibility tools gives you a much better understanding of what people are using the network for. Such tools also help you find "hot spots" where you might want to drop-in extra capacity.
Photo of Craig Mathias

Craig Mathias

  • 63 Posts
  • 0 Reply Likes
I worry that as we get to 1.8 and especially 3.5 Gbps implementations (granted, a ways out in the future), 10 Gbps will be required. But perhaps at that point 40 Gbps might be more cost-effective... And, as we move to fiber, power for AP becomes an issue. Anyway, 1 Gbps is plenty for today's 1.3 Gbps .11ac, but I don't think we can count on that being true forever...
Photo of Matthew Gast

Matthew Gast

  • 284 Posts
  • 63 Reply Likes
Even 3.5 Gbps probably is probably "only" 2.5 Gbps once you discount protocol overhead, and with a reasonable distance distribution of devices from the AP it won't all be running at the highest possible rate. I think a dual-1G link from the switch to the AP will carry us through wave 2 -- but that might only be cost effective if you ran two cables when the wiring was put in.
Photo of Matthew Gast

Matthew Gast

  • 284 Posts
  • 63 Reply Likes
802.11ac is going to be with us for several years, so having a long-term plan in mind for the wired side is important. As 802.11ac becomes more capable, the wired network will need to grow to support it. That means thinking ahead to a future with more than 1G of uplink capacity. At this point, there's not really much 10G to the edge, but ensuring that you have two 1G ports to run to AP locations is a good future-proofing exercise.
Photo of Manish Desai

Manish Desai

  • 2 Posts
  • 0 Reply Likes
For the wired side of the network, you certainly want to look into a switch solution that supports PoE+ for upcoming 802.11ac access points.

Many previous access points are getting near the traditional PoE (802.3af) standard of 15.4W of power. In order to prepare for the wave of 802.11ac, the PoE+ (802.3at) standard of 25W of power is going to be needed in order to sufficiently power next generation 802.11ac access points.

For network management, you certainly want to consider a solution that provides a unified view of wired & wireless users and provides a way of setting up unified, consistent policy enforcement for wired & wireless users.
Photo of Matthew Gast

Matthew Gast

  • 284 Posts
  • 63 Reply Likes
We also had a discussion about security in another thread: http://community.aerohive.com/aerohiv...

The short version is that there's not much to worry about for the next couple of years. Just like with 802.11n, you'll want to be using WPA2, but other than that it is fairly smooth sailing.
Photo of Nick Lowe

Nick Lowe, Official Rep

  • 2491 Posts
  • 451 Reply Likes
On the security front, it would be nice to see APs with an implementation of 802.1X-2010 / 802.1ae (MACsec) for authenticating and encrypting the backhall traffic to the wired ports.
Photo of Matthew Gast

Matthew Gast

  • 284 Posts
  • 63 Reply Likes
Getting ready for the migration to wave 2 should be something that you consider, even as part of the plans for the initial deployment. The first wave of 11ac products is useful for capacity enhancement in hot spots as determined by your network management solution. It's only in the second wave that 802.11ac is likely to become the default PHY to blanket a coverage area. You'll want to think about how to move to the second wave from the beginning though - how are you going to get sufficient power to those new APs, and how much of a switching upgrade is it going to require?