Hacker News Reader

Comment by bcoates | Hacker News Reader

Everybody hates punching holes in walls and running new cables, but this is almost certainly the wrong solution for almost every problem, outside of the weirdest of edge cases.

For starters, you're going to lose MIMO, so you've substantially downgraded your network's ability to handle multiple clients. It's also really unlikely that your weird old coax system is clean and not a rat's nest of splitters and unterminated ends (=interference).

Attaching multiple antennas to a single AP in a non-MIMO configuration is not a cost effective way of doing DAS with current technology, if it ever was. More, lower-powered APs all over the place is where you want to be spending your budget. If you want to save money on install, do PoE.

View on HN

Ethernet over power line (basically DSL) has become quite good these days, you can get fairly cheap 1gbe units today and unless you have some really crappy wiring with noisy appliances they don’t have many issues.

I looked into this pretty seriously about two weeks back when I realized that I'm losing about 2/3rds of my connection speed (30 Mbps at router) while sitting at my desk at home.

I backed off for now because I wasn't sure how well it would work in practice and honestly the speeds I do get are perfectly fine for what I do.

Curious if you or anyone else has recommendations on what I should be looking for / avoiding if I were to go that route and invest in some powerline ethernet equipment.

Can you actually find an Ethernet over power line adapter that delivers 1Gbps? From my experience you only get like a third of the advertised bandwidth.

I get pretty much 80-85% of the advertised speed on my TPLink AV1300 just over 1 gig.

With the same equipment, I get 30Mbit/s with the most convenient pair of sockets, and 90Mbit/s with a different pair.

The apartment was completely rewired in 2015.

I would say performance is so contingent on wiring quality that it cannot be reasonably assured. I tried out a TPLink system that claimed to be able to deliver 300 mb/sec and only managed 10. The fundamental problem is that you tend to want to place the units in areas with the highest potential for noise (eg near lots of electronics).

Your experience is typical even in the cleanest of AC power conditions. Powerline is hilariously oversold and regularly provides less than 25% of the advertised throughput.

This blogger has some great examples: https://specklepattern.wordpress.com/2018/03/03/seven-genera...

The "2 gigabit" adapters top out at 470mbit when both sides are plugged into the same duplex receptacle. That number is cut in half when you move to another room, and it gets worse from there.

Every Powerline vendor, without exception, are using funny math when coming up with their advertised throughput numbers. In no situation will they ever be able to provide 50% of their stated throughput, and more typically it's closer to 10%.

In this case I should consider myself lucky. I'm getting ~300Mbps out of "AV1200" adapters.

They probably go to the Shop Vacuum Power Rating school of specs. Oh you hit 40A of inrush current for 30 ms? 4500W, call it 6 HP.

They probably hit 2 gbps for 30ms on the same duplex with zero interference.

Can you give some details on the test methodology used? You are more than doubling the throughput of any publicly-posted performance tests I've seen.

If your tests show > 1gbit of throughput on a device which has a 1gbit port, something isn't adding up.

Some of those devices have multiple Ethernet ports which can be bonded.

The speed between the Ethernet over powerline nodes is independent of the Ethernet connection of the devices connected to them.

You can see the link speed between the nodes in the TPlink app.

They also support LAG, I have two of the 3 ports LAGed to a Switch.

From what I can tell the biggest issue with these other than the wiring/noise issues is if people rely on them as a switch.

I used to have 3 devices connected to them (AP, TV and Xbox) switching to LAG and a proper 1gbe TPlink switch that ironically costs more than the homeplug more or less solved my issues but ymmv.

I also think that the UK residential power wiring that uses ring circuits might be better for homeplug usecases than other curicut types.

Oh, so you're talking about the speed the adapter tells you. In my experience those speeds are about 2x what you'd get in reality over a TCP connection (using something like iperf3).

If you have time, could you run iperf3 (brew/apt install iperf3) between two computers separated by the powerline devices?

G.hn Wave 2 came out recently with a nominal 2.4 Gbps...it can probably hit 500Mbps under good conditions.

Latency. Sadly, Ethernet over powerline (IEEE 1901) made the same unfortunate choice as WiFi and the (analog phone line) modem V.4x standards in including error correction. This is not what you want or need for TCP/IP. It's better to drop a packet than to deliver it late. IEEE 1901 tries to mitigate this by using forward error correction, but succeeds in that only to a degree.

Bandwidth is alright-ish at about estimated 2MiB/s here (single pair, about 50ft apart), but RTT of ~4ms with occasional spikes in excess of 1s (at a different location, I've seen spikes in excess of 5s). This is only barely noticeable for accessing the WWW, but unpleasant for interactive GUI work (e.g. via NX or VNC) and a deal-breaker for cluster communication protocols.

1. There are still some devices designed for lan-only use that assume packet loss will be extremely rare.

2. TCP throughput suffers when non-congestion related packet-loss exceeds a fairly low value

Only an idiot would design a device to assume no packet loss, unless they were _also_ allowed to design the network. If you buy such a device and put it on a janky power network, then you have only yourself to blame.

Standards decisions should never be made to accommodate designers that want to assume there's effectively no packet loss.

Some degree of error correction is totally reasonable, but your first point is irrelevant to discussion.

To a TCP implementation, what's the difference between a dropped packet and one that's delayed by hundreds of round trip times? What problems are caused by not dropping that packet?

(If all the packets are failing to go through for that long, then you have much bigger problems than the error correction.)

And has a big probability of killing your VDSL connection, (Homeplug frequencies overlap with those used by VDSL), in addition to polluting the HAM spectrum with wideband interference.

Most houses build in the last decade or two use at least cat5 for phone lines. Just refit the outlet, plug in a switch where they all connect and, boom, wired network. Place a few strategically located wireless APs, adjust the signal strenght, and you have great coverage all over.

Edit: An alternative is to get a wifi extender that supports 5Ghz and 2.4Ghz and (this is the important part) allows you to use one of those as the back haul. You give up one of the frequencies, but you don't halve your bandwidth.

> Most houses build in the last decade or two use at least cat5 for phone lines. Just refit the outlet, plug in a switch where they all connect and, boom, wired network. Place a few strategically located wireless APs, adjust the signal strenght, and you have great coverage all over.

My parents' previous house was 2011-ish had the phone lines run over Cat5, but they still daisy chained jacks so it wasn't easy to convert to ethernet use.

That's not the only place I've seen that stupidity too.

> Edit: An alternative is to get a wifi extender that supports 5Ghz and 2.4Ghz and (this is the important part) allows you to use one of those as the back haul. You give up one of the frequencies, but you don't halve your bandwidth.

The nicer "mesh" kits in some cases even have three radios per base where the third is dedicated to backhaul.

Exactly this. I did this in a rented townhouse to get excellent signal coverage throughout while keeping my equipment locked up in a small mechanical room in the basement and requiring no modifications to the house.

You know, I never considered the unterminated end problem. Could attaching terminators to all the idle coax sockets in a house improve modem performance?

Yes. Or even disconnect them at the splitter near the ingress and terminate the splitter.

MIMO didn't really pan out as useful in 802.11ac and is still uncertain in 802.11ax. It'll be horrible for other reasons though.

Care to elaborate on that point? I was under the impression that MIMO works just fine these days.

MIMO is in common use since 802.11n, it's what "3x3" spacial streams is.

MU-MIMO, "multi-user MIMO", is what was added in 802.11ac but never gained significant support or usage. (You'll find it in some WiFi APs, but practically no WiFi clients, I've never seen one myself.)

How else am I going to afford adding WiFi to my underground lair?

Reply on news.ycombinator.com