Hacker News new | ask | show | jobs
by GuB-42 1022 days ago
The hardest part, I think, is making use of that speed.

25 GBps is about the sequential speed of a decent NVMe SSD, and much faster than anything SATA.

It is also faster than what a CPU can handle for all but the most basic of processing. Working at GBytes/s speeds require serious performance considerations when coding, often involving working with assembly, SIMD, etc... Things like the TCP stack start being a bottleneck at this point.

There is also a limit on how fast you can consume content. 25 GBit/s is about 4k@120Hz uncompressed video.

And remember that both ends need to handle the bandwidth.

Doing something useful with the full bandwidth is a challenge in itself. Running a server for many users is the most probable use case, but you need a beefy machine, maybe several to be able to serve at such speed. You need to find enough users too.

I am not saying that getting 25 Gbit/s at home is a bad thing, it is a technical achievement and I am all for it. But I feel like finding an application that can saturate that kind of bandwidth is like a continuation to the challenge. Maybe something to do with gaming or VR.

3 comments

Honestly, for most applications at home, I struggle to justify even 1 Gbit/s. At my location, I have the option to upgrade to a 1 Gbit/s connection at the click of a button. I currently get 500/500 Mbit/s for ~45€ per month. 1 Gbit/s would cost around 100€ per month.

I often think about how I could justify the increase in price for myself, but I always come to the conclusion that I just don't need it. And I am a heavy internet user. I download a lot, and I upload a lot (which is why I'm very grateful for my 500 Mbit/s in upload speed). But even with "just" 500 Mbit/s, downloading a 100 GB game takes about 30 minutes. And that's provided I have an endpoint available that can supply the full 500 Mbit/s, which is surprisingly often not the case.

If I happen to need 100 GB of data, I can wait 30 minutes.

Disclaimer: I'm talking about uplink here. Internally, between my PCs, my home server and my NAS, I'm currently looking into getting 10 Gbit/s set up. And yes, I'm sure there are some use cases out there that benefit from 1 Gbit/s uplink. But I, as a power internet user, have decided for myself that I just don't need it, which means that the average Joe definitely doesn't need it.

Still, I think it's great that it's available here and I'm all for getting that expanded further.

I think this shows a mindset difference in how we think about net connectivity vs other capacities. For things like power grid connections, car horsepowers etc, we like to size things so that we don't max them out very often.

For eg a 150 GB game download on your console, a 25G connection sounds useful.

And of course a fast connection also enables new kinds of applications when enough people have them (eg game consoles could use networked on-demand storage for assets so wouldn-t be size limited by console SSD capacities). As long as bandwidth is very unequal between users, the faster end doesn't get applications built for it. That's why we should enact policies that make fast connections widely available. Fiber is cheap after all.

The main application here is to force myself into solving those problems :) I have plans on what I'll do after. As with a 25 Gbps router comes a 25 Gbps home NAS, then I would want to have decent latency for packet processing and I hope all of that would force me to do something interesting :)