Signal Handoff Could Mean Roaming VoIP over WiFi

wassup writes "According to this article in MIT tech review (and here), researchers at University of California San Diego have developed a technology called SyncScan that will reduce handoff delay in WiFi networks to a few milliseconds. VoIP roaming will be here soon!"

Comments

What about TCP/IP handoff?

[hgilde]

Last I checked, VOIP uses TCP sockets. When you move between WiFi base stations, you first must discover your new DHCP server, then get a new local IP address, then reconnect to the VOIP server.

This will definitely be an annoying delay.

Re:What about TCP/IP handoff?

[Anonymous Coward]

Last I checked, VOIP uses TCP sockets

Ummm no, try again. They don't call it VOTCP for a reason. VOIP is a generic term for ANY technology or implementation of voice communication over IP. So it could be TCP, UDP, or even some other protocol (though this is unlikely as it would cause compatibility issues, and UDP serves the need just fine).

Furthermore, the actual voice traffic is generally transported over UDP in almost all cases.

Anyway, it would be my guess that what these guys are designing is made to workaround the traditional bottlenecks in WiFi network IP-service establishment. One thing is that when you are being handed off, the network you are being handed off too knows somewhat in advance (relatively speaking) that you're going to be handed off to it .

Re:What about TCP/IP handoff?

[c_g_hills]

This is why IPv6 is a much better network (layer 3) protocol for VoIP as it supports mobility natively, allowing TCP and UDP sessions to be maintained when roamning from one network to the next.

In fact, the Internet Society point out that IPv6 is necessary for mobile and wireless internet.

Re:What about TCP/IP handoff?

[templeton008]

sip can go over tcp or udp, it's most often used over udp though. The actual rtp audio stream goes exclusively over udp to minimize delay/complexity (it doesn't matter if some packets get dropped).

Re:What about TCP/IP handoff?

[zm]

VoIP works on UDP, not TCP. WIFI association and the address disovery can be done while the call is still in progress (ever heard of having two addresses on the same interface?). Then there is almost no impact on the media path. See for example http://sipquest.com/sipquest-software-wifi.htm.

Re:What about TCP/IP handoff?

[Alrescha]

"Last I checked, VOIP uses TCP sockets."

The last time I checked, my VoIP infrastructure used UDP.

I suspect that there are enough implementations out there now to make 'VoIP' a useless term for determining the underlying technology.

A.

Re:What about TCP/IP handoff?

[Anonymous Coward]

You are correct about the TCP/IP issues, but the design for multi-AP networks
is to have them all on the same subnet. Then the act of roaming between APs
is scoped to L2.

That leaves a couple different sources for delay/glitch:

One is if the wireless client were to set the WiFi interface "down" upon
dissassociating with the initial AP, then setting the interface back "up"
once it associates with the next AP. This may have the affect of triggering
the IP stack to release it's address information, and then restart a DHCP
transaction when the interface comes up. This sort of thing should be easy
to avoid in a WiFi VoIP client though.

The other stickler is WiFi authentication. Most WiFi users are used to WEP
with Open authentication, which is more or less instant. But a modern,
enterprise-level network with modern authentication features (802.1x/RADIUS)
makes the authentication step not-so-instant. Cisco has some proprietary
features in their APs to make roaming in such an environment a much faster
operation (they basically use an authentication "proxy" of sorts). I follow
this stuff because I'm helping develop a competing AP at the company I work for.

I'm glad to see some new ideas come out, but from looking at the article, it
doesn't look like this new research is taking the full context of the wireless
network - including the authentication scheme - into account.

Re:Cell phone?

[Anonymous Coward]

UMTS (3G GSM) which allows flexible amounts of bandwidth for a voice call has wideband (higher fidelity) codecs defined. I have heard voice encoded with these codecs and it is excellent. Note that this is possible with today's cellular radio technology without additional BW.

In addition to the slow roll out of UMTS, the issue is the cost to change all the user terminals, the sound quality the handset can provide, and that most of the other telecommunication network entities (your house phone and all the other terminals in the world) are old fashion toll quality terminals. I believe there are people researching how to convert narrowband voice to wideband voice.

The standards for these codecs and the C reference implementations can be found at 3gpp.org.

http://www.3gpp.org/specs/htmlinfo/26-series.htm

Re:WiMax will break the cell operators backs

[adolf]

Not to piss on your wet dream, but:

WiMax supports huge distances, sure. But in order to avoid needing a line of sight with the tower, you need to use low frequencies - that multiple-tens-of-GHz mumbo-jumbo is useless for penetrating things like trees, buildings, and cars.

Thankfully, old analog TV spectrum (such as the lower 700MHz band) is suitable and available for use in this way.

But realize that there's only so much information bandwidth that can be squeezed out of a slice of spectrum, and that the further you cast your signal, the more devices that are likely to be competing for that available bandwidth. So, "30-50 mile" range may not be as advantageous as you might think.

And it's all licensed and auctioned off to the highest bidder, in much the same way as existing cellular frequencies were/are. The end-user equipment doesn't even exist yet, and there's certainly no economy of scale advantage over traditional cellular phones (and there may never be, depending on how this spectrum ends up actually being used).

To top it off, it's extremely likely that the existing rules governing handheld cellular telephones to 600 mW ERP will be carried over to the lower 700's devices, if the rules aren't applicable already.

And -that- means that you'll never get more than a few miles of range. Which means again a landscape peppered with hideously expensive towers for reasonable coverage patterns.

Which is just like cellular phones operate, today. Except it's WiMax instead of CDMA.

On the other hand, it will open up the market to new players. Which will increase competition, and probably lower prices overall.

And in any event, the technology itself is not any cheaper simply by virtue of NOT being CDMA, GSM, or PCS. It's still governed by physics, the FCC, and market forces.