2013-09-14

Ubiquiti APs

I posted before about a Ruckus, and indeed, it is impressive. The only catch with a Ruckus is that to do it right you need multiple APs and a controller which soon adds to the cost.

So, I am now trying the Ubiquiti APs. There is a nice pack of 3 units which linitx do: Ubiquiti UniFi UAP AC 1300Mbps 802.11ac - 3 Pack.

High end pricing for consumer APs, but compares very well with the Ruckus pricing. I am testing the 802.11ac units which cost a bit more but still good compared to Ruckus. The more standard 802.11n and 802.11a stuff is very reasonably priced.

I had to install the app to configure them, which was easy and works on the mac with no problem. It is actually very impressive. I was able to detect the APs and attach to the network. It was very quick and easy to set up.

They are PoE, and come with PoE injectors, though we had have a couple of PoE switchs in the house. They come with wall brackets. My son may even find a shorter, white, patch lead for his games room wall, but I am not sure he cares.

The control app allows you to do a floor plan, which it will quite happily snap shot from google maps.


It has nice reports on usage, pie charts of which AP is used by which device, throughput figures, all sorts. And it just works as a seamless set of APs.

It has plenty of bells and whistles as you would expect - multiple SSIDs linked to VLANs, a "guest" management system.

What did surprise me is that it seems "better" that the Ruckus. One specific point was starting to annoy me of late with the Ruckus. I would open my MacBook Pro, put in the screen lock password, and then still have to wait a noticeable amount of time before the Mac would get a signal. Yes, this is even with the numerous amusing SSIDs all removed. I would also find that it sort of stalled occasionally when in use. The latter point was improved massively by forcing 2.4GHz, though the 5GHz was showing full signal when connected and it was close to the AP. Using the Ubiquiti the Mac has found signal before I have finished typing the screen lock password, is on 5GHz and not stalling or being odd in any way. The next real challenge will be testing VoIP calls from the iPhone over wifi, which was a strain to use on the Ruckus. So far it seems good, but I need to do a few more calls and longer calls to be sure. Obviously, even in a house, having more than one AP helps, but we had managed to place the Ruckus centrally with good signal everywhere, so that should not have been the issue. Long term I expect two will be more than adequate for the house, though I may get one of the LR (long range) units to cover the garden.

Overall I am impressed. We'll be setting up at the office as well. It does leave me with a couple of spare Ruckus to play with :-)

Why am I playing with APs? Well, we have tried many small APs, and so many are iffy in one way or another. The apple airport are around the best of the cheap ones we have found. But we get asked by customers for WiFi, and we need to really be sure what we are selling will be sensible. So we are trying different kit and evaluating at the office and in a home/office environment. We like to have used what we sell or recommend.

This does look like a system we could sensibly sell to customers and expect to just work for them. It would fit well with a FireBrick in our Office::1 package as that can handle VLANs and dish out DHCP on separate LANs for different uses.

Update: Install at office looks tidy:-

19 comments:

  1. Adrian, before you get busy with potential marketing of Unifi, drop me a line or give me a shout - we've circa 800 of them in service and are looking at having to find their replacement due to appauling performance and service and troubles with a mountain of devices and in "busy air" places.
    I can provide you with a Lenovo and Acer laptop with Realtek wireless card that will crash (freeze completely) an AP and the only recovery is a hard power cycle.

    Unfortunately the price is too good to be true for very many reasons.

    ReplyDelete
    Replies
    1. Interesting, thanks for the info. I'll see what we find as well.

      Delete
  2. Shame you didn't end up with their Golden AP, that would have made a nice surprise.

    http://www.ubnt.com/goldenAP

    ReplyDelete
  3. To be fair the realktek Wi-Fi Drivers can be rather awful, although it shouldn't crash the AP.

    I have some Standard Uni-Fi AP's kicking about somewhere That I was using at home, but I could never manage more than 65Mbit/s of throughput on them even in the same room

    Currently using Tenda Routers reconfigured to be dumb AP's @ Home get about 85 - 90Mbit/s on 2.4ghz and 130 - 150Mbit/s on 5ghz.

    I did think of plugging the Unifi's back in and having a play with the Seamless handoff they're supposed to support in firmware 3.xx beta.

    On Another note one bit of UBNT kit i've been impressed with is the Edge Router Lite, Perticullarly with the newer Alpha releases where they've got the ip acceleration working for IPv6 and Vlans.

    ReplyDelete
    Replies
    1. I had a lenovo with a realtek.. it would take about 20 reboots of the laptop to associate with the AP. The drivers are *complete* crap.

      Didn't crash the AP though.

      Delete
  4. I have several networks of these and they've been brilliant so far. I've gone with the cheap £50 Unifi AP units which have the same neat software controller and shiny interface, at less than the cost of most consumer wireless points.

    ReplyDelete
  5. For VOIP on the iPhone... What's your favourite app? I tried Bria last year on our two iPhone 4S handsets, but it seemed to drain the battery (even when set to only use Wifi).

    ReplyDelete
  6. One thing I've been curious about for a while is Wifi roaming - I have 2 bog standard(ish) consumer grade APs at home, one a WRT54GL running OpenWRT and the other a crappy Netgear thing. Both bridged onto the same VLAN, both have the same SSID, so in theory devices should roam between them (although they are running WEP for various legacy reasons)... Except every device (be it laptop, phone, tablet, etc.) seems to associate with the strongest AP when its turned on, and doesn't roam to the other AP until it completely loses the signal from the one its already associated with. The upshot is that if you're wandering around the house, at some point the AP you're associated with becomes so weak as to be useless, but the device can still see it so won't bother roaming until you turn the network interface off and on again.

    So my question is: do the big systems with a central controller (such as Ruckus) do some magic to make roaming work better, or is it just as terrible?

    ReplyDelete
    Replies
    1. My understanding is basically "yes". The controller can instigate a hand over, and that is the sort of thing they do, but I have not read enough to fully understand. What the ubiquiti stuff claims to do is some patented zero handover stuff where the device does not even realise it is changing AP and so forcing devices that hang on to an AP until it dies before re-connecting still get forced to roam. Sounds clever, but messing about here it does seem to "just work".

      Delete
    2. OK this is really interesting. After years of trying to get good coverage across the whole of our house we've settled on 3 APs each running DD-WRT, and suffering all the unpleasant edge effects of clients not handing over properly.

      Something where multiple APs cooperate and present themselves transparently to the client as a single AP would be absolutely fantasticm, and I would happily pay a significant premium for it.

      Shame that Pete is not so convinced by the reality of how these actually work...

      Delete
    3. Roaming is completely under the control of the client. Some controller-based systems have hacks that try and force clients to move AP, but not many. I've heard of some (e.g. Meru IIRC) that apparently give each client a "fake" AP to connect to, so the client doesn't realise it's moving, but this is just another hack around the fact that the controller/AP currently has no way to tell the client "there is a better AP over there, please move along".

      You can imagine the user support problems this causes. Devices are designed to stick to an AP as much as possible (think of your home, where these things are designed for and sold into - you want your device to hang on, even when the signal has practically faded away at the bottom of the garden). Contrast enterprise, when the device *should* roam whenever it has a stronger signal from another AP, but they don't. Then you have devices using up airspace way away from the AP they are connected to, interfering with the nearer AP because it's talking at full power to reach the distance - and getting slower speed (remember, longer distance = slower speed = using up more air time = other clients get less time to talk in the shared RF space). The best options here being to trim AP transmit power if possible (reduce cell size) and to cut off the slower bandwidths - so don't let APs use less than say 18-24Mbps. The AP will broadcast the SSID at the slowest data rate (which uses up lots of airtime, even if all the clients are high data rates). Trimming slow rates means that as soon as a client moves far enough away that it has to slow down to get the distance, it is unable to talk, so is forced to reassociate with another AP, or lose its connection. In higher density it's better to install a new AP and give a) that user a better experience, and b) not let that user cause problems for everyone else.

      So reducing AP cell size will help the clients decide that they need to roam, which will improve the experience for everyone.

      In terms of clients; many Windows wireless drivers have a "roaming sensitivity" type option in the driver options, where you can tell them to hang on as much as possible, or to roam quickly - which is the better option for enterprise. But user-owned devices (BYOD) won't have this setting changed, so it's not that useful. Android tends to have all sorts of problems roaming (depending on the version), and can often require turning wireless off/on when moving from one office to another. Again, BYOD brings challenges.

      The best things we can hope for seem to be 802.11k and 802.11r, which enable the controller to inform the client that it should move to a better AP, and speed up the transition. Don't let the fact they were published five years ago mislead you, though - they are only just about starting to be implemented in wireless systems (latest Apple iOS has them I believe). Both client and infrastructure need to support them, and industry tends to drag along painfully slowly in my experience.

      Summary - roaming is a client-driven function. Solutions to help for now are use a controller based hack, or trim slower bandwidths, drop TX power where AP cells overlap, and wait patiently for 11k and 11r support.

      (FWIW - I run the wireless network at a university, using Cisco controllers.)

      Delete
    4. Matthew yes I absolutely can understand why my solution of 3 consumer access points has trouble for all the reasons you discuss (though thanks for the hint about disallowing low speeds; I will try that out and see if it improves things).

      I don't know anything about what's involved in negotiating and maintaining a data link between an access point and a client under 802.11 but I have to guess that given sufficient CPU and low enough latency between two devices there's no reason why they couldn't share sufficient state between themselves such that they can hand off a wifi association already in progress without the client noticing. The timing requirements can't be that stringent since it's a lossy medium so there must always be allowance for retransmits and the like.

      As for whether this can be achieved for £200 per AP is a different question, though Ubiquiti's product info is quite clear that their product is transparent to the client (and that their network emulates a *single* AP): http://dl.ubnt.com/datasheets/unifi/UBNT_DS_Zero_Handoff_Roaming.pdf

      Would be very interested to hear more real world experiences!

      Delete
  7. We use the Mikrotik boxes, don't have any problems with them and are rock solid boxes!!

    ReplyDelete
  8. Hi, just my 2 cents: Being using Motorola AP6532 for 3-4 years now, they have an internal controller (up to 24 APs).
    Absolutely no problem, all the bells and whistles you can hope for (and more). The only problem was to buy the support package (but that was down to the box pusher (they still call themselves reseller)). Don't think they got any 802.11ac (yet).

    ReplyDelete
  9. Hey SMabille, Just wondering, does the Motorola AP6532 handle handoff and how has your experience with that been in practice?

    ReplyDelete
  10. Also, I meant to ask, RevK, I had been looking at the UniFI UAP AC but this review (http://www.smallnetbuilder.com/wireless/wireless-reviews/32153-ubiquiti-uap-ac-access-point-reviewed) amongst some other "known" issues with the AC model (no zero handoff on 5Ghz) documented on the UBNT forums had put me off them. So was thinking of UAP Pro in the meantime, had you experienced many problems?

    ReplyDelete
    Replies
    1. We got the AC for future proofing, and they seem to work fine (not zero handoff though).

      Delete
  11. Ruckus have just launched cheaper SME focussed brand: xclaimwireless.com

    I currently use Unifi at the office, but have ordered a Xi-3 to use at home.

    Looking forward to see how it performs, UBNT night have to drop their prices to compete.

    ReplyDelete

Comments are moderated purely to filter out obvious spam, but it means they may not show immediately.

Deliveries from China

I have PCBs made in China (well Hong Kong). This is all my many small PCB projects (not FireBrick). I would rather use UK suppliers but I am...