Is Wireless Roaming a Standard Feature?

[John Kotches]

This may not be the best place to ask, but is the roaming a feature of the wireless standard, so any two (or more) WiFi devices can function in this fashion?

For "ease of use", obviously same model is best but this is a generic question at this point.

 

[remixedcat]

Roaming varies from brand to brand of AP.

Typically (Cisco, Extreme, etc) all tie it into a WLAN controller and layer 3 roaming is handled that way

However APs from Cisco Meraki don't run on a typical controller based architecture. The controller is hosted in the cloud and all local layer 3 roaming is handled through a VM concentrator or a MX security appliance/router.

Setup on the Meraki units is very easy. However layer 3 roaming is a bit harder on the other two I mentioned.

 

[stevech]

This may not be the best place to ask, but is the roaming a feature of the wireless standard, so any two (or more) WiFi devices can function in this fashion?

For "ease of use", obviously same model is best but this is a generic question at this point.

Roaming...

Definition 1: Ability to automatically change from one access device to another, in the same SSID, if allowed to take many 10's of seconds, or more, and not necessarily choosing the "best" access device, is in the IEEE standard. It's merely network re-discovery - same as first-time discovery. So it's not roaming per se, IMO.

Definition 2: Ability to automatically change access devices quickly, say, a small number of seconds (but not sub-second), and reuse the security credentials, is NOT in the IEEE standards. Only enterprise grade WiFi, plus controllers and special client software, can do this. Typically Cisco, Aruba. Not found in affordable consumer WiFi. This is "fast roaming" IMO.

What consumers want is #2 above without a controller and without special and proprietary hardware/software, and with the client device choosing the BEST access device (best signal). After all these years, it's still not there for consumers.

WiFi without enterprise grade controllers, isn't good at handoffs/roaming, as cell phones do.

 

[Dreamslacker]

What consumers want is #2 above without a controller and without special and proprietary hardware/software, and with the client device choosing the BEST access device (best signal). After all these years, it's still not there for consumers.

To add, the more affordable units that do this are Ubiquiti Unifi, and surprisingly, Apple Airports. The latter doesn't advertise it but they do hand-off very rapidly once you've configured them properly.

The only caveat with the Apple units are that they generally need to be in the same generation and configuration generally requires a Mac if the units are too new (Airport configuration utility for Windows not updated as quickly).

 

[John Kotches]

Definitely looking for something along the lines of the ubiquiti and/or Apple airports.

I don't have the coin right now for the Cisco and/or Aruba devices.

I wish the Ubiquiti UAP-AC was closer to the price of the Apple Airports.



Sent from my SCH-I605 using Tapatalk

 

[azazel1024]

I think everyone else covered. But, just to provide my 2 cents of personal experiences, I have two routers, one in AP mode, one as a wireless router. Both are Netgear 3500L, Netgear firmware.

I have 4 devices as examples.

iPad 2, iPhone 5, Asus T100, HP Envy 4t with Intel 7260AC in it.

Walking from one side of the house to the other with the router and AP at opposite ends (signal strength ~-75dB on one router, -35dB on the other, and vice versa):

iPad 2 takes about 10 seconds from the time you stop at the far side of the house to switch over.

iPhone 5 takes roughly the same 10 seconds.

Asus T100 set to agressive roaming or whatever the top setting is, takes about 10 seconds to switch. Set to the default level, it takes about 12 seconds, on the lowest setting it takes about 15-30 seconds and sometimes it just doesn't want to let go of the other router.

HP Envy 4t set to agressive roaming, about 7 seconds, let in the middle about 10 seconds and set on the lowest it'll tend to hold on to the other router.

With an "agressive" test by doing it outside where signal strengths are lower, my wife iPad 2 and my iPhone have a tendency to lose network connection as signal strength drops too low on the connected router before it'll connect to the new router. The T100 and HP Envy 4t/7260 rarely lose connection before switching to the new router. So this also means that if walking very quickly between points with those two devices, sometimes it'll switch as quickly as 3-4 seconds as I walk by the "peak" signal of the nearest router and it starts dropping off as the far router starts reaching too low signal strengths.

At a jog, it'll drop the network connection before reaquiring the network at the closer router.

Roaming speed improves slightly if both routers are set to 300Mbps or 150Mbps, about 2-3s faster switch overs for all devices on the in the home test, but I keep my living room router on Channel 6+11 and 300Mbps and the basement router on Channel 1 and 150Mbps to divy up the airwaves (no nearby neighbors).

So, yeah, fast roaming, at least with my gear, doesn't happen, but its generally pretty tolerable roaming without losing connections walking around the house and even outside (though the Apple stuff will lose connections sometimes roaming outside before reacquiring).

 

[stevech]

what you're discussing is not what I call roaming. It's more like normal loss of connectivity or high error rate that triggers network re-discovery.

I think the term roaming means the change is sub-second and usually directed by the access devices. To do this takes a controller-based WiFi system. Some that are used for handheld VoIP, for devices like http://www.vocera.com/index.php/resources/videos/handling_calls?group=b3000/ handoff (roam) fast enough that VoIP buffering hides the change and the user hears no discontinuity.

 

[Dreamslacker]

@Azazel: What you're describing isn't true roaming. The connection will get terminated when the device switches between APs.

In true roaming, your device only sees one AP because all the APs adopt the same MAC as far as the device is concerned. The controller (or master in the event where any AP can fulfil this role) will determine which AP will respond to the device.

i.e. You have 2 APs and both receive the same data. They negotiate internally (or with a master controller) and the AP that is closer will respond. To your device, you just have 1 very huge AP covering the entire area.

 

[sinshiva]

not that it matters in the slightest, but i'd actually consider the zero-handoff capability to be more like a 'wireless mesh', whereas getting booted from one ap and seeking out another to be more like 'roaming'. kind of like how a roaming cellphone connects over another provider when they lose signal to their own, etc.

 

[stevech]

not that it matters in the slightest, but i'd actually consider the zero-handoff capability to be more like a 'wireless mesh', whereas getting booted from one ap and seeking out another to be more like 'roaming'. kind of like how a roaming cellphone connects over another provider when they lose signal to their own, etc.

Not really. In cellular, the term roaming means moving from a cell site owned by carrier A to one owned by company B, where A and B have roaming agreements that are both technical and financial. But moving moving among cell sites within the same company/carrier is called handoff. In cellular, very few carriers support roaming with call/connection preservation. Example: Say Sprint has a roaming agreement with a rural area carrier named zzz. Most often, when leaving coverage of Sprint, the phone must search for a new base station with which Sprint has a roaming agreement, then reconnect. It's getting better, but most often that "roam" on to zzz's network is a dropped call.

In cellular, the decision to change from base station 1 to 2, (intra-company) is usually made by the handset. Some systems do base station directed handoffs (handset is commanded to change to a particular new base station and sector/channel). CDMA cellular (Verizon and Sprint in the US) has 3G soft handoff; this is where 2+ base stations are receiving the handset's weak signal. The base stations' digital frames are voted and combined based on the bit errors in each frame. Thus, soft-handoff means that 2+ base stations' receivers are combining received data until the handset moves so that one base station is too errored and is dropped in the combinations and voting. 4G LTE has some of the same concepts.

In WiFi, the IEEE standards lack a definition of how to do handoffs as defined above. So Enterprise systems that do fast handoff do so by proprietary means and special software is needed on the client devices. Cisco started doing this with the original Aironet/Cisco hardware long ago, and generously open-sourced the protocols for the client PC and some handhelds, named Cisco Certified Extensions or some such.

The term mesh network in wireless is about routing among access nodes - rather than how a user accesses the network. The goal in a mesh is to reduce the number of "egress" points - for example a wired or wireless connection directly to a router that has a WAN connection. In wide area coverage mesh systems, it's cost prohibitive to have a lot of access devices have direct connectivity (say, cat5 or a dedicated wireless link), to reach the router/Internet. The inter-node wireless can be the same RF channel as user device are on, or can be via a second radio in each node to do this "backhaul". With the demise of Earthlink's adventures in citywide mesh networks some years back, mesh networks are mostly seen in all-on-wheels/feet military networks.

 

[sinshiva]

nice post; i understand why you'd disagree with my using 'wireless mesh', being that the ap's aren't creating point-to-point connections between eachother wirelessly. i meant more that connecting to one of these systems is more like connecting to a mesh itself rather than any one AP, and the mesh handles the specifics of Tx and Rx. whereas with a typical wireless network, the wireless client has to lose connectivity, etc. before making the decision to look for the strongest available signal and reconnect.

so in essence, i think it makes more sense to define a roaming device as one that makes the decisions itself (albeit helped along if it's being punted by the AP for low signal), whereas connecting to a zero-handoff wireless system that makes those decisions for the client be referred to as something else. client handoff, etc.

 

[remixedcat]

layer 3 roaming

 

[stevech]

so in essence, i think it makes more sense to define a roaming device as one that makes the decisions itself (albeit helped along if it's being punted by the AP for low signal), whereas connecting to a zero-handoff wireless system that makes those decisions for the client be referred to as something else. client handoff, etc.

I suggest not trying to define/redefine common industry terminology. Roaming is not in IEEE 802.11.

Host-directed handoff is a well understood term, but in WiFi, requires a controller.

Controller-less multi-AP WiFi (for SMBs) need for the client devices to learn or know all nearby APs. Cisco's good idea way back was to have each AP include in the SSID broadcast beacon a list of the nearby APs' MAC addresses. This enabled a client to learn the nearby APs without having to scan all the channels with a dwell time on each. Simple layer 2.5 software on the clients then used that list to decide which is the "best" AP to choose quickly if the current AP's signal degrades too much. This is a controller-less way to do fairly fast AP changes. I stop short of calling this hand-off because in cellular, hand-off is usually base station commanded. Roaming is not a good term either, as it usually means, to system engineers, changing systems, in WiFi terms, changing SSIDs and/or encryption credentials. And with 802.1x, a client has to re-authenticate with the AAA (RADIUS) server when changing systems.

Methods to speed all this are all proprietary due to the absence of a standard in IEEE 802.11.

And for home users with a tiny few APs, vendors don't worry with it as the whole issue comes up only for users operating while walking with a handheld and hoping there's no glitch in a video/audio stream. Hard to walk and watch without colliding with a wall.

 

[remixedcat]

Meraki does it with an MX security appliance or a VM concentrator

 

[azazel1024]

Interesting info. That probably is exactly what is happening, actual network reconnection, but at least in terms of user experience, with most things I've tried, its seamless, which I think is what matters the most.

Facetime, Skype, CIFS transfers, RDP, HTTP page loads and a couple of others I've tried just keep going without a hiccup.

 

[stevech]

Interesting info. That probably is exactly what is happening, actual network reconnection, but at least in terms of user experience, with most things I've tried, its seamless, which I think is what matters the most.

Facetime, Skype, CIFS transfers, RDP, HTTP page loads and a couple of others I've tried just keep going without a hiccup.

Many others here, over time, have not been so fortunate, using a different mix of client and access device vendors.

 

[sfx2000]

I suggest not trying to define/redefine common industry terminology. Roaming is not in IEEE 802.11.

Host-directed handoff is a well understood term, but in WiFi, requires a controller.

Controller-less multi-AP WiFi (for SMBs) need for the client devices to learn or know all nearby APs. Cisco's good idea way back was to have each AP include in the SSID broadcast beacon a list of the nearby APs' MAC addresses. This enabled a client to learn the nearby APs without having to scan all the channels with a dwell time on each. Simple layer 2.5 software on the clients then used that list to decide which is the "best" AP to choose quickly if the current AP's signal degrades too much. This is a controller-less way to do fairly fast AP changes. I stop short of calling this hand-off because in cellular, hand-off is usually base station commanded. Roaming is not a good term either, as it usually means, to system engineers, changing systems, in WiFi terms, changing SSIDs and/or encryption credentials. And with 802.1x, a client has to re-authenticate with the AAA (RADIUS) server when changing systems.

Methods to speed all this are all proprietary due to the absence of a standard in IEEE 802.11.

And for home users with a tiny few APs, vendors don't worry with it as the whole issue comes up only for users operating while walking with a handheld and hoping there's no glitch in a video/audio stream. Hard to walk and watch without colliding with a wall.

Roaming is defined in 802.11-2012 - the roll-up release.

Thing is, it's an optional feature within the spec, and many of the AP/Clients vendors have rolled out proprietary solutions.

Some require specific client SW, others are a bit more open - in other words, your mileage may vary...

First hand experience - Cisco does, but feature has to be enabled at the WLC, Aruba is similar.

In the SOHO space - first hand experience with Apple Airport - Roaming is explicitly supported in that product line for their current gear going back through at least 802.11n, and some of the older Extremes that support 802.11g do as well.

For other SOHO brands, basically need to check with the vendor first - some do, some don't.

 

[azazel1024]

Many others here, over time, have not been so fortunate, using a different mix of client and access device vendors.

It could help that I am using two of the same routers. Netgear 3500L, one as router and one in AP mode. Though the later is going south fast. The AP used to be rolling as the router, but then it stopped routing traffic over the WAN. It would work for 3-18hrs and then both WAN routing and the admin console would lock hard and I'd have to reboot it. Wifi/LAN traffic uneffected. I swapped routers functionally and it seems to work okay as just an AP, but lately I've noticed the occasional rare time when it refuses a new client connection until it is rebooted. Its only happened a couple of times...but still...

Also I seem to be having issues in the last week or so where internet page loads will stall for a bit, say 15-45s before everything works just fine for awhile. This only happens when connected through the dying AP. Lastly, for some very, very strange reason, Ping and port scanning seems to not work through that AP. I can do both through the main router over wireless, but if I try to do it when connected to the dying one, I get ping time outs and port scans fail completely.

Very bizzare, but since I know its dying, I guess not unexpected that other things would go mammory glands up.

Planning on replacing with a TP-Link Archer C7 in the next couple of months (looking like sooner rather than later now) and when I can afford it down the road, replace the main router with one as well, just to try to keep Router/AP commonality across my network.

I have DEFFINITELY seen a lot more issues when multiple vendor, or at least chipset, routers and APs have been on the same network. With roaming, general connectivity, etc, etc, ad nauseum.

 

[remixedcat]

Don't the C7's have some firmware issues relating to wireless client associations of certain devices? Did they correct?

 

[azazel1024]

Don't the C7's have some firmware issues relating to wireless client associations of certain devices? Did they correct?

IIRC it was an issue with Broadcom clients, I think specifically 802.11ac Broadcom clients. I don't happen to have any, but also I believe it was also corrected a couple of months ago.