1/13 Super interesting post (as usual) on 5G from @adriancolyer at “the morning paper”: https://blog.acolyer.org/2020/10/05/understanding-operational-5g/
“the authors investigate whether 5G as deployed in practice can live up to the hype. The short answer is no.”
But wait, there’s lots more to say.
2/ First of all, this thread is a summary of a summary. Original paper is at http://xyzhang.ucsd.edu/papers/DXu_SIGCOMM20_5Gmeasure.pdf
Title: “Understanding operational 5G: a first measurement study on its coverage, performance and energy consumption”
3/ Disclosure: I am a major 5G skeptic. Hundreds of billions in capex, no killer app.
But to address the specifics…
4/ Problem 1: The coverage wasn’t very good in their test.
Serious? Not sure, the holes were only 8% of total area and these are early days, operational experience is thin. My guess: This can be figured out.
5/ Problem 2: Throughput and latency.
Serious? Not sure, seemed TCP-specific, the proportion of net traffic that is UDP will increase monotonically as QUIC spreads through the ecosystem, and UDP numbers were good. But…
6/ …I'm getting 130Mbps right now on LTE+ in my boat. These folks got 880 on UDP but I'm really stretching to think of apps don't work at 130 but do at 880…
7/ … they mention 4K streaming. Maybe, but I'm having trouble motivating myself to replace our 10-year-old 1080p TV, which is anyhow fibre-connected. Is 4K on your phone a killer app? iPad maybe? When away from WiFi with fibre behind it?
8/ Problem 3: Application performance, “ 5G only reduced page loading times (PLT) by about 5% compared to 4G”
Serious? Maybe, but once again they ignore QUIC. However…
9/ … a lot of the latency is on-device rendering and frame processing. This might be very serious. With Moore’s law stalled (at least for single-thread processing), it’s not clear to me how much performance headroom is left in pocket-sized devices.
10/ Problem 4: Energy consumption.
Serious? Unclear, 5G design is supposed to be good at this, better radio power management *might* have a lot of headroom.
11/ Coda: “I’m intrigued by the potential for 5G to drive increasing use of edge computing platforms, getting the compute and data as close to the 5G network…” etc.
I have to confess I’m super dubious about this…
12/ Designing low-latency apps to take good advantage of edge computing is brutally difficult. Anyone know of success stories? I don't but this isn’t my specialty. I do know of wheel-spinning failures.
13/13 Pardon me for being boring, but I keep coming back to monster capex and no killer app. Interesting times…
You can follow @timbray.
Tip: mention @twtextapp on a Twitter thread with the keyword “unroll” to get a link to it.

Latest Threads Unrolled: