M2M GPRS/UMTS SIM cards: how the data traffic is calculated and charged?

On 2015-06-11 pozz wrote in comp.arch.embedded:

We had similar questions and problems some years ago, thing may have changed and there are ofcourse differences between providers.

No, you have to count all bytes you send, including the IP headers.

No, all bytes are counted, not just net IP traffic. Don't forget to count the return data...

And you can't just 'ping' a GPRS device, it has a private IP in the telco's network, not a public one. The GPRS device must initiate all transfers and you should use TCP to talk back over the socket that was set up by the device.

If you want to be able to access the devices from the outside without initation by the device, you must (let the telco) set up a 'private APN'. If that is in place, you can 'ping' the devices on their IP on that private network.

Yes and the one we used had a 10k blocksize.

But this is all from a while ago, YMMV.

--
Stef    (remove caps, dashes and .invalid from e-mail address to reply by mail) 

"This is lemma 1.1.  We start a new chapter so the numbers all go back to one." 
		-- Prof. Seager, C&O 351

Il 11/06/2015 16:17, Stef ha scritto:

One of italian mobile operator (TIM) offers IP public and fixed addresses on their M2M SIM cards.

10kB? How long is the duration of a 10kB-based connection, before starting the charge of the new 10kB-block? Or it is undefinetely long and the responsibility to disconnect is on the remote device?

How did you find those info? By tests on the field or you were able to speak with a technical guy from the operator?

My guess is that telco people still try to hang on the old convention.

In the X.25 packet switching system, there was a fixed charge for each

1..64 byte segment. Thus, a frequent pinging or Keep-Alive style activity could easily cost more compared to some not so frequent actual data transfer. Turning the connection on just prior to data transfer and off immediately after did save a lot of money.

This may also explain the stupid TCP Keep-Alive default of 2 hours.

On 2015-06-11 pozz wrote in comp.arch.embedded:

OK, that is nice. But.. This means your device is accessible by the entire internet so you must make sure your security is up to date. Will you be charged for all incoming packets as well? In that case attacks (to your device or just using your device as a 'reflector') could consume a lot of data even if you just drop the attempts.

That was per GPRS session. All sessions where rounded up 10k blocks. I don't think there was a time limit on the session duration. And yes, the device had to disconnect.

There was trial and error involved. We had a device that should run as long as possible on its batteries in case main power went out. So we shut down all non-essential stuff when not needed (And since we had that option, why not use it on main power as well?). This include the GPRS modem so we had to set up a new session every few minutes to few hours, whenever sending data was required.

When the phone bills started coming in, we realized there was a problem with that approach. ;-)

And yes, we talked with technical guys from the operator about this problem and others. Sales people are probably not the right people to answer such questions. Don't your operators have a technical guide explains al this stuff? Back then ours hadn't, but I would expect improvement over the years.

--
Stef    (remove caps, dashes and .invalid from e-mail address to reply by mail) 

Reporter:   "What would you do if you found a million dollars?" 
Yogi Berra: "If the guy was poor, I would give it back."

kB).

We have the opposite problem. The USB modem would lock up after disconnect ion. System reset is not enough. Only power cycle can restore it. Right now, we have to run something constantly to avoid disconnection, and perhap s an external watchdog timer to power cycle the device.

Fortunately, T-mobile does not charge for overage (2.5GB/month), only slow it down to 3G. We don't really need 4G anyway.