Hi again.
Thank you for your appreciation. You can hit karma + button if you would like. It is just for info, I hate the karma begging posts, but it is here so you can use it.
.
Take my comments as some kind of hints or topics for your own thinking about. Finaly, you will build and operate the network.
I would like to dedicate few lines to finance first. Especially, when you write that the connectivity will be free for everyone, there must be some financial model that will pay such "free public service". Maybe you should also shortly describe it. I am asking for that because you also have to think on the way how to cover the cost and build the network topology that would support this as much as possible. Even maybe public authority would donate you (for some time - nothing is forever), there will be some criteria that should be met. And your network should be able to meet the criteria. And, what is worse, you have to be prepaired for the moment when money pipe will be closed.
This is also linked to the black passangers problems. Did you think about what to do with locals that connect to your acesspoints with powerfull radios and overscream the mobiles? Once you provide such public wifi, I personally would be one of the firsts that will connect to your accesspoints and try to suck as much connectivity as possible.
Currently, we need to cover 3 areas like the one on the pictures above. So we have our main building where WAN link comes in, "high ground" building which will host Omnitik and to which all other locations will connect. It's the only option to have a clear line of sight. Currently, I need to cover cca 180° radius from "high ground" building to see all clients. They also aren't all in the same height. Main backbone link is in the same height, 500m away. Clients are 5-10m lower and 80-200m away.
Given the small number of clients, silent area(I think I only catch 1-2 5GHz links on high-point location, both from indoor routers) and relatively wide area to cover, do you think OmniTIK would still work for me? It's also cheaper than buying Sector+RB+mPCI wifi card. Don't get me wrong, I don't mind going sector way if it's worth it.
Well. At described situation the omnitik can work good. At least for the beginning. Maybe later, if there will be more 5GHz fixed clients, you could expand/add some panel/sector antenna to serve the most utilized sector and off-load the omnitik which still can keep other all-around widely spread clients. This is the main point for connectivity distribution, as I understood. Anyway, the final speed of traffic is really small - you wrote about 10/5MBit/s. Therefore there is no need for high throughput. I would suggest to use as narrow band as possible and as far from used frequecies as possible. You will prevent interference and keep other channels for your other connections.
I saw your way of modifying Omnitik to have POE out and it doesn't look hard to do:) I know I lose some options compared to UPA version, but then again, in case I need to reset another device connected on OmniTIK, I can just restart Omnitik, since it's power adapter is on an easily reached place.
For the first look it looks a bit easier than it is. But everyone has to try it on his own.
. Main problem is not to put too much solder on the wires but you have to boil it well. Test every contact by trying to tear it off. Of course, you do not have to solder all ports. There are also other options how to modify the poe out of omnitik, you should consider them also.
If you just restart the omnitik, it will not have influence on other devices. Only power cycling by unplugging the power supply will power cycle all connected devices at once. This is why you should also overlarge the power supply, because device booting can take much power than "normal" operation.
Can you tell me what you thought there? What should I look at here? I'm new to VLAN's over WLAN networks...
Sure. The link to wiki should show you that there is no problem with VLANs when using L3 licenced devices. Do you see? (Unlimited VLAN interfaces starting L3)
As I described earlier, my "clients" will actually be public outdoor locations in town. Equipment will be mounted on certain houses where I will have access to electricity. That is why I need a "both-side" POE injector.
Yes. This was written meanwhile or I udrestood it later. There are multiported POE injectors, you can use them. Or you can solder additional connectors to the adapter and use standard individual POE injectors included with devices. Also you can use the multiconnector fork (one female, many male connectors) instead of soldering. Anyway, you will maybe need to connect the devices together by switch (if more than two on site). Therefore it would be smart to put directly some switch there and power everything from it. I would suggest to use RB750UP for this purpose or to RB750 and solder it in the same way like omnitik. They are cheap (10-20 USD second hand) and work well. Of course, you have to put according power supply. I use old notebook supplies for this purpose. They are rated at 18-22V normally and with 3-5A output. And rock solid, with typical consumption of such place around 1-1,5A they do not heat and hold for ages.
About RB2011UIAS-RM vs. RB1100AHx2. Since this will be a public wifi, I will try to do my best to cancel all torrent traffic. We will probably start with very little limited traffic control and go on from there, depending on usage. We will definitely set up some queues for some QoS(web, voip, mail and such before other non-important traffic). But better to be safe then sorry, so why not start with a RB1100 and just be safe for the foreseeable future:)
Exactly. This will not have major influence on the budget. And as I understand your situation more and more, you will need tons of firewall rules, queues and other stuff that will slow the traffic processing. The RB1100AHx2 seems to be approximately 5times more powerfull than RB2011. RB2011 is really good for small office or home network where all clients are under control but not for such exposed public network.
As far as 2.4G coverage goes. You now have some idea of what I want to achieve on locations. When you say there can be 20 mobile clients connected to one AP, does it matter if clients make any traffic or if they idle? I mean, can there be 50 idling clients connected and 20 actually using wifi or is the limit the number of connections on it self(with or without traffic)? I can see mobile devices trying to connect on it own, even when users wont be using wifi. I so plan to set some rules to disconnect clients when their signal is too poor.
Yes, it does matter. You can see some posts here on the forum that people complain about poor wifi throughput and finally the reason is one client that makes the AP to transmit slowly and what is worse, to retransmit again and again. As I wrote before, such client has to be disconnected in order to keep others to communicate well. There is also another problem linked with this. Even the client seems to transmitt at good power, he can have problems with reception. This is something that you do not see on signal level. I have tried to solve similar situation (reverse, client had poor power, but heared well, both could happen). First I implemented the script that disconnected weakly transmitting clients. This client immediatelly tried to reconnect. As sript runs regularly, it can catch accidentally also other clients that just do not transmit with enough power because they do not need it (e.g. WMM influence, I guess or other power saving feature implanted in smartphones). So therefore I was not happy with such behaviour and I needed to serve the worse client also. So I added additional ap closer to it. Therefore I think that you should have more sectors on covered areas to allow connections of more clients if necessary and to provide them enough time for communication. And to lower the risk of problems with weak clients. Maybe also forced disconnections at certain signal level could help you in case you will have such problems (and I bet you will).
None will tell you how much mobiles will your APs hanlde. You will just try and see. But be prepaired to put additional sectors in case it will be needed.
Why would you recommend sectors instead of omni's for poor signal devices? Because beams are more focused and devices move from one antenna to another more quickly? But don't we have another problem here then, since sectors have longer range and can pick up noise from further away?
It is always a trade off. Advantage of sectors is that they do not catch interference from side or backwards. This is really huge advantage against omnis. Next, they can produce more clear signal for the clients within their range.
Just make a test. Take some AP, put it on the window of the house (similar place like you would like to use), attach an omni antena, take few mobiles (yours, friends, 3-5 will be enough) connect them to AP, run bandwidth test within 3m distance simultaneously. Then go 30m away from the window and run bandwidth test again. Maybe even 30m will be too much. Than connect some panel antena to AP and do the same. It is cheap test, maybe it will not cost you a penny, but will give you an experience. You should test more scenarios (involve 2 friends, one at 3m distance, at 10m distance and at 20m distance and test simultaneously). Looking at clients CCQ and signal quality and rates will be interresting. After that imagine that you wanted to serve dozens of mobiles within 50m range by omni! Maybe youtube would be good - but everytime run different video on all mobiles.
Now the noise of sectors. Well, you can again fight against the noise a bit. Have you seen the sector antennas of mobile operators? Have you thought about why they are tilted upwards even the clients are on the ground? In case you will have problems with noise of other APs and frequency change will not help, try to tilt the sector antenna up a bit. It will reduce the signal strenght of all devices a bit, but it will also lower the noise. May not help everytime, but could be usefull to know and try it. This cannot be done with omni.
Other thing. If you use groove, it has only one chain. Many mobiles have two chains today and of course they are never perfectly aligned towards your antena, so using dual polarity 2x2 mimo AP will be also very beneficial. Therefoe SXT G-2HnD or QRT-2 could be your candidates. But they have gigabit ethernet, and I am in dobts if they would accept 100mbit POE from RB750... Also basebox would be an option, but it needs a pigtail and an antenna and has gigabit also. Maybe the 100mbit POE extractor and gigabit POE injector could be the solution for gigabit devices powered from RB750, or multiPOE or so. Just an idea, I have not made it yet.
I have one desire. Woud it be possible that you can share with me (personally or publicly here on the forum, if you do not mind) how you proceed with network building, including the problem descriptions and their solutions you will face? And pictures, of course?