MGig Musings

It used to be a relatively common statement to say “your wired connection will always serve you better than your wireless connection”.  With AC Wave 2 though wireless connection speeds and quality have now surpassed GB Ethernet.  Pretty amazing really - utilizing AC Wave 2 in a proper configuration and ideal environment you are better served using a wireless connection than a wired ethernet connection!

Note that I said “ideal”.  All the marketing material from all the different vendors (that obviously have a vested interest in propagating the technology) will promote blazing fast speeds, etc.  But I consider those blazing fast speeds to be kind of like the zero percent interest rates advertised at the car yards.   Sounds good - but are you gonna get it?   

I’ll talk about AC Wave 2 in a different post though.  Right now I’m going to share my thoughts on what AC Wave 2 means to the backend infrastructure.

Right now we have a lot of companies telling us that in order to support the blazing fast speeds of AC Wave 2 and the added net traffic that it is going to create we need to address the potential limitations of how the access point is connecting to the network.  Specifically the limitations of traditional GB Ethernet.  If you have multiple clients connecting to a single AP at the (theoretical) maximums of 1.3gbps/7gbps, dependent on frequency, then that single 1 gbps ethernet connection from the AP to the switch simply isn’t going to cut it.

Makes sense.  Lets explore that.  Lets assume that we do have multiple clients connecting at those speeds.  Or, rather, lets assume that we have, at least, so much demand on a single AP that a 1gbps connection from AP to switch is going to get overloaded.  We need to do something right?

That’s where MGig comes in.  MGig, or MultiGigabit, is a new tech that allows a single cat 5e, 6, or 6a cable to transport more than 1gbps.  Up to a theoretical (there’s that word again!) 10 gbps actually.  

Awesome.  10gbps.  Lets say we are conservative and have 5 AP’s per switch.  50gbps over wireless per switch.  Doesn’t quite meet my needs if everyone is connecting at 7gbps but it’s better than nothing.  That was a close call……I better get MGig everywhere, on every POE switch, in every closet, so that I am future proofed.  

Thank the networking gods that MGig came along.  

Quick Caveat:

Despite how my comments below may appear - I do actually like MGig.  I think it is a good solution to address a need, assuming that the need is there.  I am not suggesting not getting it.  And I am not criticizing anyone that has blanketed their environment with it.  I’m, rather, suggesting that if you are considering it that you take a step back from the “state of fear” that a lot of people seem to fall victim to and take a realistic look at its benefit, its potential, and your environment, and then make decisions from there.

And again - the comments below are based on everything operating at its theoretical maximum, which is a highly unlikely scenario anyway.  But its a good exercise.

Ok - now lets talk reality. 

Bottlenecks.

You’re going to have network bottlenecks.  Put 50gbps or more into a switch and unless you have 50gbps everywhere else - your uplinks, your WAN connection, your servers, your Internet connection, etc - then at some point that 50gbps is going to hit a wall.

For WAN environments it is safe to assume that the majority of the traffic at the satellite sites needs to go to, at least, the core/hub site.  At least in education environments.  So take a look at those links.  Are they going to support the high volume of traffic that you are supposedly going to generate?  Probably not.

Let’s say you don’t have a WAN need and all your traffic is local; Localized servers, etc.  Are your servers going to support that traffic?  Possibly - that can be done.  A decent datacenter can receive multiple 100gbps feeds.  But those are very specialized environments.

And what about your Internet connection?  In edu environments especially, the large majority of network traffic is now Internet bound.  Many institutions are moving everything offsite.  We have.  In our district 90% of what we do is now Internet based - Google Apps, hosted LMS’s, online testing, etc, etc.  Very little of our network data services are hosted locally anymore.  And that doesn’t even consider the erroneous traffic (YouTube, etc).  Assuming that I have 10 switches with 5 APs each, all clients running at the theoretical max so that each AP is hitting 10gbps each - that is 500gbps.

I don't think it is reasonable to, at least yet, assume anyone reading this has a 500gbps Internet pipe.  Probably more in the range of 1gbps.  Maybe 5.  

That's not to say that you won't go that high.  But while "future proofing" is important it is also equally as important to consider the here, now, and near future when making purchases.

These are big numbers of course - and unrealistic - but the point is take a look at your network as a whole.  If you are considering implementing MGig then consider the path that all of that traffic will be taking.  If you have a satellite site with a 100mbps connection back to your central site that hosts all your servers then how much benefit is the ability to connect your AP’s to the local switch at 2.5-10gbps really going to bring you?

POE Limitations.

Just to reiterate.  I like MGig.  I think it is awesome.  But when I was looking at solutions I was surprised, and disappointed really, to see the POE limitations on the switches that incorporate it.  In most cases I found  power availability reduction of around 25%.  That means that there is 25% less power available for AP’s.  And that means that you can connect 25% fewer AP’s on a MGig switch than a traditional POE+ or UPOE switch.  They support POE+ and UPOE of course - it’s just that the available power appears to be less.  At least from my research.

It's a trade off for increasing the bandwidth but for environments with a lot of AP’s - that’s kind of a big deal.  That means you’re may actually need more switches than you currently have if you want everything connecting via MGig.  

More switches to accommodate that also potentially means a lot of new, unused, ethernet ports which is kind of counterintuitive when trying to boost your wireless capabilities.

AP Reduction…..Are You Sure???

“But with AC Wave 2 we won’t need as many AP’s - so the need for more switches won’t exist!!!”.

Maybe - but that is VERY dependent on your environment.  Beam forming and all that good stuff means that your AC connected devices will perform better over similar range than N, and the MU-MIMO capabilities of Wave 2 will stomp all over AC's feeble MIMO, and if they can all connect to one AP then sure.......you might not need as many AP’s.  

But that doesn’t account for signal interference - walls, etc - client thresholds, and so on.  In schools, especially old ones, there are some pretty heavy duty interferers that just come with the territory.  Regardless of how it looks on paper at a high level, you just might simply need to maintain the same number of AP’s that you currently have in order to retain signal quality/strength where you want it.

But lets say you can reduce the number of AP's while maintaining adequate coverage.  What about the switches themselves?  Having a lot of people connect to one AP might be desired - and the switch port may theoretically accommodate the volume of traffic.  But that is a heck of a lot of packets per second (pps).  Don’t forget that switch ports have upper limits not only on the volume of data they can transport but also the number of pps they can accommodate.  Consolidate multiple inbound MGig switch ports and push out that traffic to a single WAN link or upstream switch and you might run into pps threshold issues.  That is something that people often overlook.  Admittedly it is hard to reach that max but it's worthwhile considering it.  You don't want to make a large purchase and then get caught out by something like that. 

And lastly there is the realistic capability of multiple MGig ports per switch.  A switch may have 12 ports that can accommodate MGig connections but the ASIC for those ports, at least on the switches I looked at, is shared.  They are not individual line speed uplink ports with direct access to the switch fabric.  They won’t all operate at max capacity simultaneously.  It’s a shared backbone.  Not that needing such a thing is even a realistic probability but if you are thinking that you can rely on 1/2 a dozen or so AC Wave 2 AP’s all in the same MGig capable switch to support your entire environment make sure to take a close look at the potential volume of traffic per port and see if the shared ASIC will support that total.  You might be able to spread your AP’s out amongst ports so that the ASIC distribution/utilization is equal - but if you really want to be able to max out everything at the same time then the reality is that you won’t be able to do it AND use all 12 MGig ports.

Summary.

It’s good.  MGig is cool.  It can address very real needs.  But approach with care.  I won’t say caution because I don’t think that it is something that will disappear - just care.  If you have someone telling you that you need it everywhere then make sure to really take a look at your environment to determine for yourself if that is the case.  You might very well benefit from putting MGig everywhere but only you can make that call.  When you step back, take a breath, and look at the path for your traffic, potential bottlenecks, AP placement and effectiveness, etc, you might find that you don’t need it quite as much as what you thought.

My recommendation?  

I don’t think it is a bad thing to get SOME MGig into your environment now.  We can all talk on how we think things will turn out in the next few years but the truth is........stuff changes.  Something better might come along.  Alternatively MGig might also be as necessary as electricity (I couldn't think of a more creative analogy) in the future.

I don’t think a bit of “future proofing” is a bad thing.  We don’t have a need for MGig in our environment right now but, considering the reasonably low additional cost to do it on a single switch, I have no problem with ensuring that a few of the replacement switches I buy are MGig capable.  But only in strategic locations - ones that have the potential for such a need like larger gathering places; gymnasiums, theaters, etc.  That way if the need does arise unexpectedly then something is in place for it.  And if the need pops up in another unplanned location……simple……just move that switch to that location.

But do I think that we should have MGig on every switch in our environment that supports connected AP’s?  No.  Not at this point.  I figure if we take the due diligence course and have a smattering of MGig capable switches throughout the environment then we are prepared.  If the need arises in years to come to spread that capability further then we can address it at that time.  

If you're reading this you are probably involved in schools and like most schools, funds are limited and it is up to you to make the wisest investment - trying to find the happy medium between need and expenditure while making decisions that won't leave your environment in the stone age.  To that end, for larger scale deployments, I always think it is wisest to let the need, not a vendors prediction, drive the investment.  

I’m not willing to make a large investment to make MGig happen now.  But I am ok with a smaller investment “just in case”.