1.4.b Architect indoor and outdoor RF deployments
There are some overarching considerations when designing for any of the scenarios below. The list can get quite extensive so if I had to pick the 5 most useful I’d probably go with:
- What minimum speeds are required?
- What are the roaming requirements (if any)?
- What environment am I deploying in?
- What client/device density are you expecting?
- What types of devices are going to be used?
Each one has its own list of sub questions but just starting with the above should provide a good idea of what sort of design is required.
Minimum speed will alter the cell size immediately as in general the higher the minimum required speed the smaller the cell size.
Roaming requirements may not matter for a general deployment but if Voice over WiFi (VoWiFi) is the plan then suddenly the parameters got a lot stricter.
Band Support is dictated by the devices in use with 99% of devices supporting the 2.4 GHz band and therefore greater distances. Fantastic until you have two APs on the same channel in close proximity which then need to share the airtime with each other, causing all devices associated to both WAPs to suffer and lower their throughput. Exacerbated by only 3 non-overlapping channels in most countries (Japan, 4) means you’ll see most techs pushing 5 GHz as aggressively as possible. The crux being 5 GHz only will reduce your coverage levels.
Modulation is probably one of the most overlooked components of WiFi but is a good indicator of interference. When a device connects to an AP it has already taken note of what protocols and speeds are supported/mandatory within the SSIDs becaon frames and, based on its own proprietary algorithms, it will attempt to send data at what it believes is the most appropriate modulation rate on the most appropriate band (if the SSID is dual band). The example below shows an OSX device and its behind-the-scenes settings which reveal much more than a 5 bar signal indicator. This can achieved through pressing the alt button and clicking the WiFi icon in OSX.
- 802.11n Protocol
- 5 GHz Band on Channel 36
- Transmit rate of 450 Mbps
- Modulation Coding Scheme (MCS) of 23
If you have a look at wikipedia or in fact http://mcsindex.com/ you can directly link modulation rate of 23 to a throughput of 450Mbps. The modulation in use is 64-QAM 5/6 which the best possible modulation when using 3 spatial streams, 40MHz channel width and Short Guard Interval (SGI) which is a protection mechanism used within 802.11n. In a nutshell that’s what you’ll get with a single client sitting 1 M from the AP and no interference (yay). A solid indicator of distance is the RSSI which in this example is -57 so I’d actually guess the client is maybe a bit further away.. lets call it 5 M.
The point of modulation is your device may be reporting a cracking signal (5 full bars!) but your network is still running like a dog for some reason. You then decide to do an alt + click on your OSX device and notice something like this.
The device is still running 802.11n which is a good sign and the signal strength is still very good at -63 dB but your rate is 52 Mbps and MCS Index is 11 (16-QAM 1/2). Using the MCS index I can tell the connection is only using 2 spatial streams and is limited to 20 MHz channel width. Very understandable considering it is in the 2.4 GHz range but that knowledge isn’t going to help much when you’re trying to copy (purchased, of course) HD movies across the network.
In all honesty 52 Mbps isn’t horrific for web browsing but lets assume your housemate just turned on the microwave and you see it drop to and MCS Index of 2. Sure the theoretical rate is 19.5 Mbps but halve that because of the half-duplex medium and chuck in some re-transmissions because of that microwave and thing aren’t looking good for that HD Game of Thrones episode you just
I went completely off topic there but the point is you can be close to the AP and have a great signal but still have a bad connection. This is important during surveys or troubleshooting as you can use a spectrum analyser to determine interference and where things are likely to go awry.
For windows devices there is a way to find out similar information on your connection but it takes a bit more nous and I forget the lines every damn time. A chap called WiFi Nigel has already done the hard work:
- 1.4.b [i] Coverage
Coverage is easy; set all your APs to 1 Mbps mandatory rate, full power and done! Not quite. This will ensure RF coverage to the far nooks and crannies of your building whilst achieving at least an ICMP response from the servers but that’s all you’re going to get. By allowing lower rates/protocols you’ve now allowed a device in those far reaches to be attached to a WAP and lowering the cell rate for all associated devices to that lowest speed. In the example below, one laptop is sitting within the higher rate range (not all rates are shown) whilst the second laptop has had to lower its rate due to high re-transmissions caused by distance/interference. In a deployment where 802.11b rates (1, 2, 5.5, 11) are disabled this won’t have been an issue as the 2nd laptop will de-associate from the AP after dropping below the minimum required rate.
In this scenario you now have at least one laptop not only connected at 5.5 Mbps and lowering the overall cell speed for all clients associated to that AP, but it is also using the 802.11b protocol which further compounds the issue by enforcing various legacy protection mechanisms. Now imagine there are 10 laptops connected; they would all be forced to run at the lowest rate (5.5) and support a protocol no one wants to see in a modern wireless deployment.
In addition to raw theoretical speeds, removing an entire protocol like 802.11b means wireless clients on that network don’t have to tune their NICs to support legacy technologies and access methods which is greatly improved in each subsequent ‘upgrade’ to 802.11g/n and now ac (5 GHz only).
The key to a good coverage deployment then, is finding the balance between highest minimum mandatory speed whilst ensuring you’re not being too strict and causing the cell sizes to shrink to a point where devices can’t associate unless located next to an AP. Cell overlap for roaming may not be a consideration but the idea is to have coverage everywhere whilst maintaining a usable connection.
Adding in the next layer of complication is the band support. For a coverage only model where you’ve pumped up the power levels, 2.4 GHz will likely reach those corners but your 5 GHz network will only be reaching part of the way. This may not be such a bad thing as your furthest reaching devices can remain on the slower, patchier 802.11b/2.4 GHz band whilst the 5 GHz band can be optimised for throughput.
- 1.4.b [ii] Throughput
Throughput in its most basic form is the raw theoretical rate your clients are attached to the AP at. With the previous examples in mind it should be easy to work out that clients further from the AP are likely to have lower speeds and modulation rates therefore less throughput. Once again cell size comes in to play with the balance between high throughput and good coverage at odds with each other. This kind of deployment, depending on budget, may see some coverage holes across the buildings as the RF parameters are tuned to ensure the best modulation rates are negotiated and as a result, the best throughput.
Using the 2.4 GHz band for throughput is going to be limited. You’re likely to never channel bond in this band so your highest theoretical rate maxes out at 216.7 Mbps. Not too shabby but that’s assuming a short guard interval which is less likely in the 2.4 Ghz band and of course, you’re standing next to the AP with no interference blah blah. Realistically your best throughput will come from the 5 GHz band and so your next challenge will be tweaking that to suit your needs. Some decisions for consideration:
- Implement 802.11ac?
- If yes, what channel width – 40, 80, 160 MHz?
- Is 802.11ac Wave 2 an option?
- Do your devices support it?
- How many channels are available for your use in this band?
- Have you enabled all of them?
- Do your devices support all of them?
- What is the minimum rate you want all your devices to connect at?
What you’ll likely end up with is a much higher density of APs installed but with a corresponding smaller cell size meaning higher rates are negotiated at all times. Costly but effective as long as the channel plan is well managed, particularly with 2.4 GHz which I would encourage turning off completely unless it is required.
- 1.4.b [iii] Voice
Voice! So many considerations there’s even a separate CCNP exam on it. Oh and its own design guide too:
Voice is all about good quality signal, no interference, and fast roaming making it arguably the most sensitive of deployment models.
Cisco recommends using the 5 GHz band for voice deployments and expects a minimum RSSI of -67 at all times, particularly when roaming at the cell edge where, if I remember correctly, a 20% cell overlap is recommended to ensure a smooth roam. In addition to strong signal, 25 dB minimum SNR is expected in order to ensure a packet error of less than 1%.
Naturally this is not going to go amazingly well with 2.4 GHz so I’m immediately ruling that out as an option unless you truly love spending hours troubleshooting RF Ghosts across the buildings. Oddly enough Cisco recommends 802.11a as the recommended protocol (802.11g is 2.4 GHz only) and a 24 Mbps mandatory rate as minimum which is fairly high when you consider 802.11a caps out at 54 Mbps.
Fast roaming of some kind is definitely recommended, ideally 802.11r. There is an assumption this SSID won’t be used by anything other than voice so enabling 802.11r shouldn’t be an issue for other devices as there won’t be any. However, this could depend upon the phones in use (Cisco, of course) and your deployment type (WLC, FlexConnect, Autonomous).
Enabling 802.11k will reduce the power consumption on the phones as they don’t need to poll for neighbours whilst improving the roaming experience so that one is a no brainer!
Your AP density is certainly going to be high with this deployment but the overall TCO compared to a wired voice network is a pittance.
- 1.4.b [iv] Location
I have not had the pleasure of having to design for location so far but I can imagine it would be an interesting challenge, particularly if I were trying to retrofit to an existing deployment.
Location can be sensitive to various parameters as well with a recommended -75 dB RSSI at all times. Not quite like voice but the placement of APs is the real difficulty. With a traditional office deployment you might run your APs in a linear fashion down the middle of a floor in order to provide 360° coverage.
With location you’re expected to triangulate signals and this becomes pretty difficult with a traditional deployment. This could lead to much greater expense as you re-cable areas and surround them with wall mounted, semi-directional APs as well as a few more ceiling mounted APs to assist with the triangulation. Cost comes into play of course so working out the cheapest deployment whilst not sacrificing accuracy is the challenge and that may stem from the requirements laid out prior to the project undertaking.
Whilst this part is brief, location is complex so I recommend reading the link above and then checking out passive vs active RFID tags.
- 1.4.b [v] HD
None of these deployments are easy and High Density is no exception. In an odd way, it is probably a little easier to design for something sensitive like Voice where you are using specific devices on specific bands versus designing for HD Wireless which could require you to support a mixture of laptops and tablets in the same room on different bands and all with a consistently good throughput. Yikes.
Most of the considerations have actually been covered of the past 2 or 3 posts including this one and it comes down to planning your channels, RF parameters and AP Placements off the requirements set out early on in the project and your site survey results. This is where the planning and expensive surveys pay off as you persuade clients to join