Dean Bubley

Handset OS Fragmentation is Here to Stay

Baba

Of course arbitrage is not a networking term. It's a general concept in business and commerce, whereby *users* or *customers* can play off differential pricing of what is essentially the same good or service.

It applies to handsets where there are multiple ways for the *user* to choose between to achieve a specific objective, be it making a phone call or downloading/sideloadin... an MP3 file. For a music download, realtime is (usually) not important, so the user can look at exploiting the differential pricing between an operator-mediated portal and transferring a file from a PC.

The *user* or the *user's software agent* is starting to call the shots, facilitated by ever-more powerful devices which when combined with IP networks & the Internet can decouple access from service.

Increasingly, real-time services are becoming less important than non real-time capabilities in terms of the perception of user value (and also payment). This is already true in the fixed line/Internet world, with the exception of voice. But to deliver scalable realtime voice it clearly is *not necessary* to have IMS. Existing circuit services, over-the-top VoIP, and standalone operator-managed SIP VoIP app servers work well enough already and scale.

In any case, we are already starting to see voice calls fragment between QoS-essential (999 / 911 calls, important B2B communications, medical etc) and QoS-optional calls (phoning a mate in Australia for an hour's chat). There's no point "wasting" QoS and network resources - consumers already know this, which is why they use Skype or SMS or VoIP callthrough instead of needlessly expensive cellphone calls.

Real time mobile video is near-irrelevant. There's no massmarket business model in mobile, and I haven't seen anything that even remotely demonstrates that this is likely to change. Video should be treated as an add-on, not a core design objective for mobile-centric NGNs.

Conversely, non-realtime applications - SMS, web access, email, music/content downloads, social networks, filesharing and so on - are becoming proportionately more valuable in mobile.

*Therefore* future network investments will start to become more *optimised* for non-realtime capabilities. Realtime will still be valuable, but over time it will become secondary. This in turn will drive network capex and handset architecture decisions.

Sure, we will probably always see some form of QoS-managed network for the stuff which absolutely, positively has to be realtime. But it won't drive the overall investment decisions & certainly not define the underlying architecture for the majority of traffic for which that would be over-engineered.

This is why various of the radio-access network evolutions are evolving to some form of split-access mechanism. The stuff that matters goes to the QoS-managed operator core. The stuff for which best-effort is good-enough gets piped straight out to the Internet.

And in many cases it will be the handset (& its software & above all the user) that defines any instance of communication in terms of whether or not QoS or realtime is important.

Dean