back to article Operators demand smartphones sort signalling storm

While operators spent 2009 fretting about data traffic overloading their fragile 3G networks, this year they have also been worrying about another stress on their systems, exploding signalling burdens from 'chatty' devices that constantly poll the network in applications like social networking updates. Nokia Siemens has thrown …


This topic is closed for new posts.
Anonymous Coward

The real answer

Is a new version of HTTP that allows push. What we are seeing is a perversion of protocols created by a perversion of economics... if you bought your own phone, the network operators wouldn't see it as theirs, and the protocols would change, not the phones...



Can't be really be bothered to login properly and explain this but...

But this post is so full of fail, miss understanding or comprehensions its untrue.

Classic example - Fast Dormancy - talked about one point as being the resolution, when clearly its actually the problem, femto standards - I am guessing your talking about cell PCH or URU PCH but difficult to understand but both (and other flavours) are out there in the macro networks now.

Capacity problems talked about in 2009 are the same problem as this, it was just that no one in the publishing / media / etc could understand anything more complex than bandwidth so miss information was rife. Easy to make a bunch of false assumptions than to try and understand complex inter dependancy between / within network traffic models...

yadda yadda yadda


chatter boxes

The chatter problem (as described by a cellular engineer to me) is that the networks moving to LTE use IPv6 on top of legacy protocols. From the IP-layer point of view (HTTP and others) the network is always connected. The signal layers under IP have far too many wrapper layers each doing their own setup and teardowns. Sometimes on a per-packet basis. Sometimes on timeouts regardless of what usage is still going through. This results in a HUGE amount of sub-chatter even halfway through a regular TCP link for any network which has not transitioned properly or completely.

Websites seem to also have this fascination with hundreds of requests per page (forced to not cache and forced to close immediately) and you start to see why people are starting to really hate them.

Gold badge

polling and slow control channel

@VoodooTrucker, (and somewhat @Amos) I agree that the root problem is all this polling instead of using a push system. But HTTP? People seem to forget that these devices already have TCP and UDP on them, I think using that directly would be better. Amos, these extra layers mainly affect the latency more than anything, along with some unnecessary retransmissions (a temporary delay may be interpreted by TCP as a dropped packet which it retransmits), the big problem here is the constant polling.

Example -- my aircard and phone both go into dormant state after something like 30 or 45 seconds of no traffic. This is fine, BUT, these facebook and twitter apps, etc., will POLL for updates like every minute. Result? Instead of like now, where the card either stays active for quite a while (if I'm hopping from web page to web page, or downloading, or streaming videos or music) or it goes active for a few seconds, loads the page, then goes idle for minutes on end while I type or read (or stays dormant for hours on end while I'm not using it at all), this polling makes it switch idle->active->idle->... like every 30 seconds, all day and night.

The second part of the problem? 1X is 144kbps, EVDO is 3.1mbps, EDGE is 220kbps, 3G is, well, whatever speed depending on what's deployed, 1.8mbps (I hope there's still not plain 384kbps UMTS still around) up to 21mbps -- but the control channel is 4800 or 9600bps (that is, 0.4 or 0.9kbps). Texts go over this channel, the signalling for your phone to make a call or to be told to ring because there's an incoming call, and all this data session teardown/tearup traffic, all shares this channel. The speed on it is so low because the phone has to monitor this particular channel all the time, it's to save battery power and to help prevent dropouts (a brief dropout on a call or data session is a nuisance, on the control channel it means your phone wouldn't ring if you got a call then.)

It sounds like this solution should cut the traffic use in half, instead of the card explicitly saying it's going idle, the card and network will KNOW it's going idle if no traffic goes over the channel in a certain length of time.

This topic is closed for new posts.


Biting the hand that feeds IT © 1998–2017