The Empire Strikes Back: Or AT&Ts Micro Cell Strategy Pt2--Response from ATT
blindreason.org
blindreason.org
When AT&T is telling you not to use your femtocell for data, it isn't just because it puts an extra strain on their network, its also significantly slower than if you just used the wifi you probably already have to begin with. The extra hops it has to go through - from the femtocell, through your ISP, to AT&T's SGSN, GGSN, then out to AT&T's ISP, make it seem silly to want to use a femtocell for data.
In addition, I don't see any reason they can't charge you for the femtocell, even though it is using your ISP's resources you already paid for, it still adds value to your service.
The SIP gateways that providers like T-Mobile use seem like a better alternative to femtocell's anyways. With these, your voice and data both go over wifi, and only the voice will enter the core network. When you need to be paged on a mobile termination, the core network knows you are registered as a SIP guy, so it just sends you a SIP INVITE which reaches you over your wifi. When you roam back into the cell coverage area, it goes back to normal.
This is a classic bait and switch, as far as I can tell.
If it's the latter - then it's nuts that they are charging against your data cap for usage that doesn't even traverse their network.
If it's the former, (which I suspect it is), whether it's just or not, they'd be within their right to charge for the transit load on their systems.
UE<->NodeB (Uu): <your data> | PDCP | RLC | MAC | <L1 radio stuff>
NodeB<->RNC (Iub): <your data> | PDCP | RLC | MAC | FP | <L2 stuff depending on IP/ATM/etc>
RNC <-> SGSN (Iu): <your data> | GTP | UDP | IP | <L2 stuff>
SGSN <-> GGSN (Gn): <your data> | GTP | UDP | IP | <L2 stuff>
GGSN <-> Internet (Gi): <your data> (finally!)
At some points your data's UDP/TCP headers might be compressed/uncompressed but thats the basic idea. Something needs to perform the functions performed by all these nodes. "Traditionally" the femtocell is a combination of the RNC/NodeB (aka the Radio Access Network).
Clearly it is possible to add core network functionality to the femtocell, but not only is it probably not worth it, the security implications are tricky (you would have to expose your HLR directly to the internet to be able to talk to the femtocell - instead of just your SGSN, which you would likely have special CN nodes for talking to femtocells).
So it is almost guaranteed to be the former. And it makes AT&T's recommendations make alot more sense. You don't even need to assume any malice in them.
http://en.wikipedia.org/wiki/GPRS_Core_Network http://en.wikipedia.org/wiki/Network_switching_subsystem
... OK, makes some sense now.
Their recommendations make sense. It's actually nice of them to recommend using WiFi where available (though you'd hope people would be smart enough to figure that out on their own.)
Looking at it from a bottleneck/financial perspective I wonder where the constraints are that cause them to want/need to limit data usage. If it's just the UE<->NodeB (Uu) bandwidth, then the femtocell addresses that and maybe they could prorate the data coming through it, if it's farther up the string then it's understandable.