Hacker News Reader

Comment by rcarmo | Hacker News Reader

This is nicely done, and takes me back to when I worked in mobile network planning.

Most people are blissfully unaware of how much radio chatter and constant adjustment is actually holding up their browsing session from second to second, with the UE negotiating the best possible terms (power, throughput, etc.) with adjacent cell towers before it decides (or is prompted) to hand-over to a new cell.

Optimizing cell hand-over and adjacency matrices is something that operators need to do quite frequently as traffic patterns and network topology change and a computationally neat problem that can be scaled horizontally (I once optimized one such job to run in 4 hours instead of 24).

View on HN

I learned that the cellular protocol stack wasn't as simple as it seemed when I was debugging my VoIP library under different network conditions. I would go into the *#*#INFO#*#* menu on an Android phone and force it to a particular network type. 2G/EDGE provided just barely enough throughput for me to fit my packets through on my lowest bitrate. But sometimes there were periods when the packets would just stop coming through. This would last for a perceptible amount of time, sometimes even more than a second, but after that, all the packets I sent during that time would come through in a burst. That's how I found out about the RLC protocol that's somewhere in the stack in 2G, 3G, and 4G, and that it has this "acknowledged mode", and that in this mode it does retransmissions if there are bit errors. This also made it very clear why TCP sometimes struggles so much when running over a 2G connection.

Anyway, I wish there was something like Wireshark to which I could connect a phone and see it talk to the cell towers in real time. This would really help understand how this all works.

Um, I'm pretty sure that there is not much of that handover stuff happening while browsing hacker news since that's likely in a low-velocity scenario, i.e., sitting on the toilet.

It can also happen due to other phones in the tower area moving and needing more or less power, which might result in "the system" deciding to move your phone over to another tower to reduce congestion. Even if you are just sitting on the toilet.

You would be surprised. If you're ever using your phone with a single available cell at low signal quality, just about anything will affect your signal... how you hold your phone, which room you're in, which way you're facing, etc.

The ability to switch to a different tower papers over all sorts of moment-to-moment connectivity issues.

Might be sitting on the toilet in a train

>Most people are blissfully unaware of how much radio chatter and constant adjustment is actually holding up their browsing session from second to second, with the UE negotiating the best possible terms (power, throughput, etc.) with adjacent cell towers before it decides (or is prompted) to hand-over to a new cell.

My experience in my networking class in college and how complex even a TCP/IP connection is over a reliable wired connection was enough to make me run screaming from the field as a profession.

The fact that I can get 200/mbit on my phone today still astounds me.

It's an interesting, but even to me an intimidating field.

There is no decision at the terminal side, there is no negotiation. Network collects reporting from the terminal(s) and decide centrally.

Reply on news.ycombinator.com