Somewhere around 2004, a set of vendors came up with a specification called 'Unlicensed Mobile Access' (UMA) . The problem that was being solved was this:
There is currently no standard mechanism that bridges the unlicensed spectrum (Wifi and related) to the carrier mobile network. There was a need to bring 'mobile services' to the unlicensed spectrum, especially for phones that could do VoIP while on WLAN/ethernet and switch to the mobile operator network when out of range. In addition to being able to call both from unlicensed spectrum and licensed mobile spectrum, it was necessary that the same services tha were configured in the mobile operator network was also available when I was connected via a broadband access point so that people calling me would not know the difference (and even the caller would not feel the difference between the two different networks, from a service availability perspective.)
So UMA was born. It is important to understand that is is an access technology and all it does is that it provides a bridge between unlicensed spectrum and the licensed spectrum of the mobile operator.
It is pretty much a no-brainer - this was just the bridge that operators and OEMs needed to make the dual-mode handset dream a reality. Infact, when it was first introduced, 3GPP IMS was still quite a distance away and operators and OEMs did not have a choice but to incorporate this technology. Infact, 3GPP release 6 also added UMA as an access technology. Hurray, just incase you still had phones that did not speak IMS by the time release 6 hits the market !
However, things have changed. Over the past two years, IMS has made giant leaps in deployment and the network infrastructure needed to make IMS a reality has steadily come into shape. In fact, as we speak today ,IMS is actively being trialed by several operators including Sprint, SBC, Verizon, mm02, Vodafone, BT and several others. (so as you may have guessed, I have a different opinion from what my colleague ElusiveCheese posted here - which is good - what fun is technology if we all agree all the time !)
At the same time, several OEMs are also putting in IMS ready software in their phones to be able to receive the services that IMS delivers (I am actually working with some of them, though not sure if their plans are public - google around and you will find public ones too).
Unfortunately for the UMA proponents, IMS progressed much faster than what the UMA folks thought would be possible and the delta time frame between UMA ready infrastructure and IMS ready infrastructure is less than a year apart - which is not much at all.
This brings to the front several disadvantages of UMA that were previously swept under the carpet because there was no other choice. Some of those are:
- UMA only works for GSM/GPRS networks. Bad luck, UMTS
- Even though UMA does handover between GSM/GPRS and a Wifi network, it does not handle access point - access point handover. Which means if a user roams from hotspot to hotspot UMA cannot continue WLAN based voice calling - once you step out of the first hotspot, UMA switches the call to GSM/GPRS. In other words, UMA is a good solution inside your house not bigger. Why is this a big deal ? Well, with the advent of metro mesh networks (wifi hotspots mounted on top of lamposts by cool companies such as Tropos and others such as Cisco), it is critical that access point - acess point handover is handled.
- UMA is a tunnel, or a bridge. The phone which connects to the UMA bridge is not an entity in the end-end network. The last entity the network sees is the UMA gateway. Why is this bad ? Well, this mean that important services like presence, which involve end-end availability will not work (at least transparently) with a phone behind a UNC.
- One of the fundamental design principles of scalable architecture is this: eliminate as many inter-protocol gateways as you can. Gateways between multiple protocols are a single point of failure (just as SIP-PSTN gateways are too). They are a necessary evil when two network cannot talk directly to each other. The problem is that if the gateway fails, even if you have two perfectly operating network, they are not going to communicate.
On the other hand, if client OEM vendors directly incorporate SIP into their end points, several things are automatically addressed:
- The phone itself becomes an addressable entity in the network (so services like presence work)
- There are no limitations on access technology - IMS works on GSM/GPRS/UMTS and similar. So as you add/change transport infrastructure, you don't need heavy investments (in UMA you do)
This still leaves one issue: SIP/IMS to GSM/CDMA/GPRS handover. The requirements are still the same in terms of handover duration. There are companies actively working on this technology and from what I understand, good progress is being made. Do note that this does not still affect end -end IMS deployment in a big way. There is a heavy push from mobile operators to introduce high-speed bandwidth technologies such as HSDPA/HSUPA which they hope will make consumers use the mobile carrier network exclusively in favour of unlicensed spectrum (till they figure out how to make money off that network too ;-) )
So the big question we need to ask is:
Is there any motivation for Operators /OEMs to continue with UMA, especially since UMA too has just rolled out when IMS deployments is less than a year away ? Even if they do put both SIP/IMS and UMA, how long will UMA last ?
UMA proponents see this risk, and talk about how 'UMA accelerates IMS deployment' - in a way they are right, till OEM clients support IMS/SIP, UMA does connect the two. But I hope they (KinetoWireless and others) realize that as IMS deplyoment sets in, they had better change their business plan to move away from UMA too !