Sufficient Velocity Down?

[gatewaewanderer]

Cloudflare has Published a blog on the Outage

https://blog.cloudflare.com/cloudflare-outage-on-june-21-2022/

Apparently there was an issue when an update hit certain servers, causing


And there are graphics that show how, despite only 4% of the networks locations going down, it cause a 50% outage in network requests. I would think they should consider spreading the load out more.

Click to shrink...

Cloudflare's name has come up a few times over the last few years for crap like this, though this is the worst I can remember hearing about. Considering that, I honestly have to wonder both why more people and companies don't turn elsewhere for what Cloudflare does and more broadly, how Cloudflare has managed to stay in business in general.

 

[Czlyydwr Llrngwl]

Well, there is the point that a handful of incidents, generally resolved within fifteen minutes to halff an hour, in several years is not actually a bad record for glitches and remediation. Then too, economies of scale most definitely apply for internet hosting, so splitting things off to make the structure as a whole more robust is disincentivized unless someone comes along to make them, and that takes political will that is generally not very strong if it's present at all for internet infrastructure despite how central it's become to life in most of the most influential parts of the world. That's probably an angle better taken to the threads I'm sure SB or SV have for this, though, not QQ where current politics is specifically a forbidden topic.

I definitely agree that getting back to a more grid-like overall internet architecture rather than the central hubs and trunks that it has been steadily shifting toward for decades would be preferable, for sure. Hard to route around damage to one node if a majority, or even just a large fraction, of traffic can only be handled by that node, whether because there's literally no other path or just because the alternatives can't deal with the throughput.