I'll tell you about one of my projects, from the beginning of design to implementation. The article does not claim to be the ultimate truth, I will be glad to hear constructive criticism addressed to me.
The events described in this article took place about two years ago. The case began with the fact that one company approached us with a request to modernize one of its partially open storage warehouses, mostly unheated hangars about 7-8 meters high, if my memory serves me right, and with a total area of about 50,000 square meters. The customer already has a controller with a dozen access points. The service for which the wireless network is designed are data collection terminals that exchange information with a WMS server. About 150 terminals for the entire wireless network. Low client density and minimal bandwidth and latency requirements. The material stored in the warehouse, to put it mildly, is unfriendly to the signal: when passing through one row of products, it attenuates as if it passes through several load-bearing walls.The height of the products is at least 4 meters, if not more.
Antenna selection
It was decided to use directional antennas in order to reduce the number of access points, their mutual influence, to cover more area. The use of horned access points would not have helped due to the fact that the ceiling height was much greater than the distance between the rows, TPC with all the consequences. And it was necessary to organize the coverage in a row, because through a four-meter wall of products on both sides of the row, the signal is very attenuated, and the only chance to raise at least some kind of network is to install access points in the line of sight of the client.
Range selection
We decided to use 2.4 GHz as the operating range. Perhaps this decision caused genuine bewilderment among the pundits, and they stopped reading the post from this place, but this range was more suitable for our goal: to cover a large area with the required bandwidth at a minimum and low client density. In addition, our facility was located outside the city, it was something like a free economic zone, where other large factories and warehouses were located at a decent distance from each other (fencing, checkpoint, all the cases ...). So the problem of utilizing the 2.4 GHz channel was not as acute as if we were in the city center.
Model selection
Next, it was necessary to determine the model and form factor of the access point. We chose between points 27/28 + 2566 or street point 1562D with a built-in directional antenna. 1562 won in price, in antenna gain and in ease of installation, and we chose it. So, 80% of the access points were 1562D, but somewhere we still used omni points to “patch” various pockets and connections between the corridors. We took one point per corridor, two points per corridor in the case of long corridors. Of course, this approach simply did not give a damn about the recommendations about the symmetry of the power of the access point and clients in order to avoid the consequences in the form of one-way audibility, but in my defense I can say that audibility was two-way and the data that we needed went freely.Both during the tests and during the pilot, this scheme proved to be quite good in the light of our specific task.
Drawing up a specification
The specification was assembled, a coverage map was drawn and sent to the customer for approval. They had questions, we answered them and they kind of gave the go-ahead.
Then a request comes in with a request to reduce the cost of the solution. In general, this often happens, especially with relatively large projects. This happens for two reasons: either the customer says that he has enough money, as if he wants to and pricks, or many vendors and integrators participate in the competition for the implementation of the project and the price gives your company a competitive advantage. Then a scene takes place like a Martian in a movie: the ship must fly, but it is too heavy and then the equipment, provisions, life support system, casing are thrown away, and as a result, the person flies on almost the same stool with a jet engine. As a result, at the third or fourth iteration, you catch yourself thinking that you look like a boy from a Soviet cartoon who mixes dough with wood and throws it into the oven with the words: "And so it will do."
This time, thank God, there was only one iteration. We borrowed an access point with an antenna from the distributors and went for an examination. In fact, the search for the equipment itself for the survey is a separate conversation. For the fairness of the test results you need a specific model, but sometimes it is not, especially in a tight timeframe and you choose the lesser of two evils: either nothing or at least some equipment with dancing with a tambourine, using your imagination and calculating the trajectory of the ship's flight from Earth to Jupiter ... We came to the customer, deployed the equipment and made measurements. As a result, it was decided that it was possible to painlessly reduce the number of points by 30%.
Further, the final specification, technical specification is agreed and an order is made for a batch of equipment from the vendor. In fact, these approvals of various specifications and various details may take more than a month or two, sometimes it can take up to a year. But in this case, this stage passed relatively quickly.
Further, we learn that the delivery time is delayed, as there is a shortage of components at the factory. This eats up the margin of time that we have prepared for a leisurely setting with a break for cookies and thinking about the structure of the universe, so as not to set everything up in a hurry and not make a bunch of mistakes as a result. As a result, it turns out that there is exactly a week between the date of delivery of the project and the arrival of the equipment. That is, in a week you need to do the network setup and installation.
Mounting
Then the equipment arrives and the installers take over. But since they are primarily installers and are not required to know about the nuances of the propagation of an electromagnetic signal, you write for them a short guide on how points can be hung, and how not, etc.
Since the access points that we chose are outdoor, sometimes they come in bridge mode, depending on the nuances in the specification, and in this state they do not connect to the controller. To do this, go to the console of each point and manually change the mode. That is what we planned to do before giving all the points to the installers. But as usual, the deadlines are on, a fully working network was needed yesterday, and we just started picking boxes with a barcode scanner. In general, we decided to hang it like this. Then we recorded the poppies of all access points and added them to the MAC filter on the controller. The points were connected, the mode was changed to local via the WEB GUI of the controller.
Debugging network and access points
We hung up all the access points, about 80 in total. Of these, there are no 16 points on the controller, and there are only two points to knock on the controller. We figured out the points that the join requests did not send. There were two access points that, due to a bug, could not connect to the controller, because they could not download the firmware, because they could not decrypt the discovery response from the controller. We replaced them with spare access points. The radio of one access point was in a down state due to lack of power, we did not have access points of this model in stock, because the specification was cut, so we had to decide something.
We replaced the Chinese switch, from which only the first four ports were supplied with power to the cisco switch, and everything worked. Similar actions were required with another Chinese, since one of the ports simply did not work on it. After we put all the access points in order, we immediately found holes in the coverage. It turned out that some of the access points were mixed up during installation. They put it in place. Further, there were problems with client roaming. We tweaked the coverage hole detection and optimized roaming settings and the problem went away.
Controller setup
The deferral advisory notice has been released on the current version of the customer's controller. When upgrading the controller firmware, the old controller firmware remains on the controller and becomes emergency firmware. For this reason, the controller was flashed twice for the most stable firmware in order to "overwrite" the old firmware with bugs. Next, we connected the old and new controllers in an ON SSO pair. It turned out, of course, not immediately.
So, the project is ready. We delivered it on time, the customer accepted it. At that time, the project was significant for me, added experience, knowledge to my piggy bank and left a lot of positive emotions and memories.