Real world 4G vs. 5G test shows 14x bandwidth increase • Hacker News Board

Re: Real world 4G vs. 5G test shows 14x bandwidth increase

#91

Post by Retric » Thu, Feb 15, 2018, 9:41 PM UTC

askvictor wrote: post #90

martinald wrote: post #82

Earlier quoted context omitted.

Most sites have fibre backhaul now. And then point to point wireless can do 10gig+ these days anyway to a tower that does have fibre backhaul.

Even with fibre, it wouldn't take many users on the theoretical capacity of 5g to saturate the uplink.

5g is shared bandwidth from ~20Gbps download 10gbps upload, increasing the number of users over a minimum does not increase peak bandwidth just utilization.

Considering multiple carriors would each have 20gbps and there are ~200,000 cell towers in the US we are talking massive bandwidth in aggregate.

Re: Real world 4G vs. 5G test shows 14x bandwidth increase

#92

Post by jsgo » Thu, Feb 15, 2018, 9:54 PM UTC

woah wrote: post #65

jsgo wrote: post #31

Earlier quoted context omitted.

I'm genuinely hoping that it puts ISPs in a bad position to have no dominance in the market. I've been under a "provider" who caps out at 10Mbps with, sadly, about 40% uptime with our neighborhood for about 5 years (constant outages. On a first name/you've got my phone number basis with the contracted technician, as is quite a few people in the neighborhood). I joined the board to get out of it. We're due to get out…

What's the difference between an ISP and a cell phone company?

I can switch to any cell phone provider I wish at any given time. ISPs that isn't the case for the vast majority of the US.

I was originally with Verizon, but the pricing was a bit steep so I left (grandfathered unlimited data plan, but was paying pretty heavily for 2 said plans).

I switched to Sprint and that was working fine, their unlimited data plan was working and the prices were much cheaper. However, the previously mentioned fiasco with our ISP meant that tethering was now important to me as a fallback when needed and Sprint's plan didn't include tethering and it was a bit steep at the time (as a developer, I have to work on things at night sometimes from home and garbage internet won't suffice there. I empathize heavily with the people in the neighborhood who work from home constantly/online classes, I do).

So I switched to T-Mobile with the Simple Choice plan at the time with basically same features, but 14GB tethering per line and added two more lines for the kids at a reasonable rate. January I switched to the One plan and added International Plus (unlimited tethering) and I predominantly tether now and only connect to my actual router that leverages the garbage ISP when I need to use my internal network (and I check the internet at that point and it is still bad). Plus, tethering is typically 3x faster than said ISP. Needless to say, I absolutely love T-Mobile in our location, it isn't great everywhere though from what I hear, and it works fine for me as an outright replacement (I don't do Netflix or other streaming from computer, so my tethering isn't particularly damaging. Plus, not in an area of high congestion).

So there's a ton of difference here: I've been able to move around as needed to get what works best for me when it comes to the cell phone company side of the equation and I've been hamstrung with the ISP side of things. Plus, I don't even dislike Verizon and Sprint, they work well, but it just didn't work for my needs. With ISPs and the limited competition, they have little incentive from their peers to make me like/accept them, case in point the ISP we're stuck with until end of March.

Re: Real world 4G vs. 5G test shows 14x bandwidth increase

#93

Post by ilovecars2 » Thu, Feb 15, 2018, 11:47 PM UTC

PeterisP wrote: post #80

jfindley wrote: post #71

Earlier quoted context omitted.

I would really like details on latency. Bandwidth is all very well, but it's high latency that is so often the real limiting factor in browsing performance. I live in a rural area, and my wired connection is very slow (~ 12Mbps). 4G competes well for bandwidth, but the latency (and the fact that the latency is highly variable) makes 4G unsuitable for general purpose internet use. If 5G could be made closer to WiFi fo…

Isn't the latency mainly determined by congestion / bandwidth saturation on either the radio network or operator's backbone? IIRC the actual technological latency of a 4G link is just ~2 ms extra; the problem is that you have to share the link with many people, so if the network is overloaded you may get unstable latency. In this regard, anything that increases capacity would automatically help with latency as well.

Yes, but also retransmission that happens at the physical/link layer to mask lossy links from higher layer protocols, such as TCP.

Re: Real world 4G vs. 5G test shows 14x bandwidth increase

#94

Post by ilovecars2 » Thu, Feb 15, 2018, 11:49 PM UTC

jfindley wrote: post #79

iknowstuff wrote: post #77

Earlier quoted context omitted.

In cities, I routinely get pings of about 40-50ms on LTE. I think it's well past the point of matching fixed broadband in that regard.

Interesting. Here (UK) fixed broadband is much better than that for latency. I ususually see 8-11ms on fixed broadband (DSL-based), yet my ping on 4G is generally more like 40-150ms, with lots of spikes.

Even rural broadband tends to average 30ms pings for idle connections. 4G, like you say, is massively variable.

Re: Real world 4G vs. 5G test shows 14x bandwidth increase

#95

Post by ilovecars2 » Thu, Feb 15, 2018, 11:59 PM UTC

martinald wrote: post #84

xur17 wrote: post #58

Earlier quoted context omitted.

Would this transparent handoff work for just calls, or for data as well? I'd love to see a solution to the wifi to 4g data handoff issues I run into.

Multipath TCP is a solution to this, Apple has it in iOS. But adoption is really low.

There are also some unsolved challenges with Multipath TCP, at least last time I studied it in 2014/2015.

For example, when multiple good links are available, Multipath TCP will tend to flip flap between them rather than stabilise on one or the other. You can also get massive reordering which will tend to throw off the congestion control algorithm, but I believe that issue has since been resolved. Good chance are there are some other issues that make MPTCP not quite suitable for widespread use right now.