VDSL vs Fiber? Practical differences and latency?

Hi,

this is a bit off-topic, but since many people here use their OpenWrt devices for routing, I thought some might have experience in this regard.

I'm wondering whether it's worth upgrading my internet connection to fiber (FTTH) as it will become available in a couple of weeks at my home. Currently, I use a VDSL (PPPoE) connection.

Now, obviously, I can calculate how much faster down- and uploads would be at different bandwidths. But I'm more interested in the practical differences during daily tasks, especially in terms of latency. Bandwidth is actually not a priority to us because 250Mbps down & 40Mbps up are already plenty for our use (web browsing, streaming, working remotely, but no gaming). More upstream would be nice from time to time, but it's not crucial. So, the only advantage of fiber over DSL left would be latency, I suppose.

I'm wondering: Do you notice a difference between a fiber and a VDSL/PPPoE connection during daily browsing? Does fiber feel "snappier"? Or is the difference of a couple of milliseconds just not noticeable?

For reference: My ping times to the PPPoE peer are currently 8-9ms and to servers of major services or CDNs like Google and Cloudflare 15-17 ms.

One more note: As for the PPPoE overhead: There wouldn't be a difference in this regard with the fiber connection because the provider here uses PPPoE for their fiber connections as well.

Thanks and regards,

Timo

Latency depends more on the ISP, than the technology (assuming reasonable baseline speeds). Getting rid of PPPoE encapsulation would be beneficial, though.

1 Like

With a stable DSL link this is not so obvious, but fiber promises massive improvements in resistance against interference. For DSL it can be enough if one self or a neighbor starts using a power line (PLC) device to experience nasty interference issues ("can be enough" not "will be enough", PLC can also coexist peacefully with DSL if one is lucky).

Latency differences will be typically really small, VDSL is clocked at 4kHz so the lowest theoretical one-way latency over the DSL link is capped at 0.25 milliseconds (in practice it might take a few clocks) in GPON the frame duration is 0.125ms (GPON is faster so a 125µs GPON frame has higher payload than a 250µs DSL clock). So theoretically they are different, but practically not to a degree that will matter for most applications (and not at all for applications suitable over the internet with its considerably larger RTTs).

Probably not noticeable if all else is the same...

These are mostly dominated by the ISP internal routing and peering/transit points with your target services. That can theoretically differ between DSL and FTTH (ISPs are free to do what ever they want in their backbone networks) but there is no reason this has to be different.

The current DSL rate 250/40 and the fixation on PPPoE independent of link technology makes me wonder whether your ISP is Deutsche Telekom :wink:

+1; but if my guess is correct not very likely.

3 Likes

Thanks @slh and @moeller0, especially for the detailed reply. Sounds like it wouldn't make much of a difference for the given use case then.

I haven't had any issues with the performance and reliability of the line over the last years, so let's knock on wood it stays that way :wink: In any case, I'm not planning to use powerline and I can only hope, our neighbor won't either.

Close enough. My ISP is actually 1&1, but they simply use the network/line of Deutsche Telekom. The new FTTH network is also built by Telekom and 1&1 recently started reselling/marketing those links as well. So, yep, PPPoE is here to stay – at least for me.

1 Like

Yes and no ;). They do use Telekom's last mile access, but they either buy the complete package from Telekom (WIA) so you get an IP address from Telekom's range/pool and Telekom's uneven peering, or, if they already have fiber access to the BNG your link terminates on they use layer2 bitstream access (L2-BSA) in which case your IP address is from 1&1/Versatel (you can use whois to figure out who owns your IP address) and you also get 1&1/Versatel peering.
They are also the only other big competitor that currently re-sells Telekom FTTH (and apparently all over WIA and not L2-BSA yet?).

And yes, Telekom has PPPoE backed-into their whole-sale platform and the regulatory approved standard contract offers, so getting rid of PPPoE will take a while (if it ever happens). There are worse things :wink:

Worse things perhaps, but it's bad enough for the typical ftth speeds, as that implies needing much faster hardware (and/or proprietary acceleration) to cope with the PPPoE CPU overhead than otherwise required.

2 Likes

That is how proprietary ISP routers work for some years now... the core CPUs are often to puny to allow the required packet massaging at "modern" access rates so heavy lifting is delegated to offload engines that often restrict the generality (which does not matter much for ISP routers as these typically only expose limited capabilities the user). I am on a fence on this, on the one hand I am unhappy about the loss of generality and configurability but on the other hand offloads often help to save considerable amounts of energy....

I used to have fttc and ping times of something like 20ms and now rely on LTE with double that. I don't think I can tell the difference even with that extreme.

I'll bet with a blind test many could not.

It's a bit like back in the old days watching a crappy quality pirated film. Did the quality difference really matter? One's mind makes up for a lot and takes time to process stuff anyway.

For sure I noticed the difference between 56k and ADSL. That was a big change. Albeit I miss the handshake noise - the excitement of dialing up and gaining access is gone. In the UK there was an 0800 free to call ISP and you had to keep trying to get on and you'd be cut off after an hour. After repeated attempts the sense of excitement was terrific.

Only if the SOC is slow and designed on an old process (admittedly, this is perfectly describing most common routers). Atom J1900 can easily idle at 6 watts and can cope with these tasks at (1000BASE-T) wire-speed, you won't drive them much beyond 20-25 watts under (artificially-) high load. And while my currenty ivy-bridge c1037u based router is not a shining example of conserving power, it's totally unimpressed with these tasks (at 15 watts).

So yes, it isn't that hard to cope with these demands, on the right platform.

Ah, interesting. I've read about that difference before, but didn't know that the ip address hints to the difference. I have an ip address form 1&1/Versatel, so that would be L2-BSA then. Which would also mean that if I ordered fiber now, I'd get the peering of Telekom (whether that's better or worse, I don't know, even though I've mostly read complaints about Telekom's peering).

Guess I am lucky that my turris omnia is plenty powerful and frugal at my 116/37 Mbps link with CPU cycles to spare for an IDS (used to track usage by internal host and volume). I bought two raspberry pi's sequentially and always wanted to use those as wired-only routers.... but I never got to it, the raspberry pi400 does duty as light desktop for one of my kids the rpi4b got converted into a GPS-disciplined NTP server so I have quality time available for benchmarking... but I digress.

Well, I already use a PC Engines APU2 as my primary router, so I guess that should suffice :slight_smile: It's at least much faster than most of the consumer devices used with OpenWrt, I guess.

Routing at 1 GBit/s, yes - but not much more in addition to that (PPPoE might already get tight, SQM, not at that speed on these tired jaguar cores).

1 Like

I've read some review/benchmark for the APU2 that said, it can do 1GBit/s with PPPoE on Linux, but not BSD (PfSense/OPNSense). But even if I switched to fiber, I wouldn't get a 1Gbit/s plan at the moment, because I simply have no need for these download rates. I'd probably go with a 500/100 plan, mostly because of the increased upstream compared to the VDSL offering. But then again, while faster uploads would be nice some times, it's not too often that I have larger files to upload either.

But after reading that the latency and day to day use wouldn't really change in my case, I'll most likely just stick with VDSL for the time being.

I switched off my bevy (gaggle? flight?) of APU for routing and firewall as they didn't have enough to do much more than 500 Mbps NAT and routing†. I use them now for jail servers for DHCP, DNS, mail, NextCloud, web, ... Somewhere, in the deep, distant past, I did some speed tests of ath79, ipq40xx, and amd64, including the APU2.

† "More" meaning more functionality, such as SQM or packet inspection such as Snort.

I've got a GL.iNet MV1000 "Brume" I just dusted off for guest-network duty. It is running a bit over 3 W at idle. I don't recall how well it handles throughput, nor have I kept up with the DSA and NAT-offloading changes over the last few years.

The Hardkernel H2 was a great option with "Intel(R) Celeron(R) J4115 CPU @ 1.80GHz" and the ability to take an Intel 4-port server card over PCIe (dual, bonded interfaces). Unfortunately, supply-chain issues have resulted in its discontinuance.

amd64 with (hopefully) genuine 4-port Intel "pulls" is my current choice, though I don't know which compact, low-power option I'd buy these days.

Edit: I haven't read it or thought much about it in the last three years: Comparative Throughput Testing Including NAT, SQM, WireGuard, and OpenVPN

1 Like