This summer I had an opportunity to do a total redesign, which was my first complete WLAN design from scratch. One of my schools was having some renovations done so we recabled and got rid of the hallway-fi. This post will go over the steps I took during my design process, look at a few of the things I’ll do in the future as part of the WLAN life cycle (Plan, Deploy, Validate, Monitor, Rinse, Repeat), and some things I learned in the process.
For any signal strength heat maps you see, colours indicate -72 dBm or higher; gray is between -72 dBm and -85 dBm; white is -85 dBm or lower. For channel overlap maps, green indicates no overlapping access points, and yellow indicates two.
Step 1: Plan
After tracking down a twenty year old floor plan, I set to making it current, using an advanced CAD tool called Paint. Because my design had to be done before the renovations were finished, I met with our maintenance department about what exactly the new areas would look like. I then completely forgot to add one new wall into the floor plan. Next it was off to the Andrew Von Nagy’s handy-dandy capacity planner. As part of the IT department, I knew what our requirements were. If this was a consulting gig, now is the time I’d spend some quality time with the customer going over their requirements, clients, applications, planned growth, locations, etc. You can combine the results of the capacity plan with the floor plan to determine where you might have groups of users, isolated offices, high-density areas, or other special situations. Here are my capacity requirements.
The next thing I did was get on-site to measure the RF attenuation of the various types of wall construction in the school. I’m not going to explain the process in this post, but here are a couple of great resources on measuring loss:
- Wi-Fi Planning, Walls and dB’s – Measuring Obstruction Losses for WLAN Predictive Modelling guest post on Ekahau’s blog by Nigel Bowden (blog, Twitter)
- Predictive Model Accuracy Enhancement webinar by Fluke Networks with Devin Akin (blog, Twitter)
I found five different construction types:
- Drywall (3 dB)
- Cinder block (8 db)
- Cinder block with locker (10 dB)
- Double cinder block (16 dB)
- Exterior brick wall (10 dB)
Armed with three key ingredients of a WLAN design (floor plan, requirements, RF measurements), I was ready to get down and dirty with Ekahau Site Survey. Unfortunately ESS’s Wall Outlining Wizard doesn’t support JPEGs, so I had to spend some time drawing walls and coverage areas. The wall that I forgot to move was only one small drywall-type wall. Here is what my initial design looked like.
The missing wall is located between the staff room and nursery school room. It really was a silly error on my part – it was demolished before I did my site visit, and then rebuilt before I did the install. I’ll add it in before my next survey.
I used transmit powers of 6 dBm and 12 dBm for 2.4 GHz and 5 GHz respectively. I’m now thinking those were too low – we’ll get to my validation survey shortly. You may have noticed, if you counted, that I have nine access points, and only one 2.4 GHz radio is turned off. I’ll address these two points after we look at the validation results. I should also note that low rates are disabled, with 24 Mbps the minimum basic rate and beacon rate.
I didn’t count the gym AP into the capacity plan because it’s only used for special events. It’s there for presenters or parents at sporting events. If ever there is an event that promises to bring in more active clients than one AP can handle, I’ll address it as needed. It does need to be part of the channel plan, as it will be heard in some places outside the gym.
I did get on site again and do some AP-on-a-stick measurements. Here are a couple images to show the comparison of results from my predictive design to my APOS measurements.
There are some not-so-subtle differences. I honestly don’t think I did the APOS using the transmit powers that I ended up using in the deployment, and I can’t remember what they were. Another thing I did wrong: my AP-on-a-stick was more of an AP-on-an-eight-foot-step-ladder. All in all, not optimal and possibly a waste of time. Note to self: be better next time.
Step 2: Deploy
We had electricians do all the cabling on this job, which was nice (and saved about a month). We still spent a week terminating, testing and mounting. We ended up with around 140 drops. I mounted all the APs myself, taking care to put them in the correct locations and orient them properly. Not much to tell.
This was also my first opportunity to use a patch antenna. I used an AccelTex dual-band 4/7 dBi gain model attached to a Juniper WLA-532E. Here’s what that looks like. I really like the mount that they supply, with the double articulating bracket and pipe-mount option.
Step 3: Validate
I finally got a chance to do the validation survey, after waiting for antennas to arrive and for my schedule to open up. Here are the results.
I ended up with some channel overlap that I didn’t expect; however, none of the overlapping channels are primary coverage. I also got slightly better signal strength in some areas where it wasn’t needed, and a massive hole where I wasn’t expecting one (second classroom from left). I’ve got some changes planned – namely, shutting off a radio or three and increasing transmit power in a couple places to fill in coverage. With that will come a revamp of the channel plan. There are quite a few laptops in this school that are 2.4-only, so I need to be make sure they can access the network everywhere. I’m also going to make sure the floor plan is correct before surveying again. Now, on to step 4.
Step 4: Monitor
I’ve been keeping an eye on the client balance between 2.4 GHz and 5 GHz. Here’s what is looks like over the course of a day.
Nothing like colour consistency in an NMS eh? And I took those screencaps on two different computers. This school has 79 2.4-only devices out of 119 total, so the 802.11na devices are either the 18 iPads, 22 of our newest dual-band laptops, or personal devices on our public SSID. I’m still trying to sort out a way of seeing retries in RingMaster, but I am not holding out much hope. I also need to go do some testing with an iPad and a 2.4-only laptop.
So, I’m going to make some tweaks and survey again, and see where we end up. I’ve heard somewhere that WLAN design is an iterative process.
Things I Learned:
- Make the process shorter, or make better notes. Three months is too long for nine access points!
- Do validation survey with classroom doors closed.
- Make the APOS survey, if done, more accurate by using correct TX powers and AP placement (see first point).
- Remember to put ALL the walls in ESS.
- Think higher transmit powers, less radios; this is not HD.
Thanks for reading!