Noticing when an app is only hosted in us-east-1 • Hacker News Board

Re: Noticing when an app is only hosted in us-east-1

#191

Post by onion2k » Fri, Jun 30, 2023, 8:51 AM UTC

phamilton wrote: post #185

onion2k wrote: post #54

Earlier quoted context omitted.

Those 104ms of latency don't make a dent. Aren't there several requests to complete though? Assuming 5 synchronous requests before the first byte of your web page is rendered (a DNS lookup, a TLS connection to establish (which requires 2 requests), then a preflight request, and finally the actual request) that's a full half a second just on the flight time regardless of all the other costs. That's an extra 10% on top…

Just a nit, but aren't DNS lookups going to be independent of distance to us-east-1?

That depends on where the name server is. If a website is only hosted on us-east-1 it's common to see the canonical name server being in the US as well.

Re: Noticing when an app is only hosted in us-east-1

#192

Post by arghwhat » Tue, Jul 04, 2023, 11:12 AM UTC

yuuuuuuuu wrote: post #190

arghwhat wrote: post #73

Earlier quoted context omitted.

HN takes - when fully cached, where it only takes 1 request - ~200ms to load, so 3/4ths of the pageload is just the roundtrip. HN user experience in EU is "meh" because HN is one of the fastest loading dynamic pages on the web to load, meant to give you whiplash from the load speed. Not all pages can load with 1 request, even when designed well, so they will be sub-"meh". HN, if hosted in EU, would have gotten a "daa…

I think you are missin the point. The typical website is so bloated that a latency of a single roundtrip is not the root cause. They may load quicker if roundtrips are faster, but that's then mainly because they make too many roundtrips. That it makes a significant difference for HN (as in: "measurable") is because HN is so snappy in the first place. Apart from that, being located in Europe, I agree with the other po…

Latency of a single roundtrip is the sole reason for the bloat being an issue. The bandwidth requirements to load even the heaviest webpages at a few megabytes is negligible for most people, but these pages are slow to load even in gigabit connections exactly due to their many render-blocking round-trips chained together.

Re: Noticing when an app is only hosted in us-east-1

#193

Post by arghwhat » Tue, Jul 04, 2023, 11:15 AM UTC

TacticalCoder wrote: post #126

arghwhat wrote: post #73

Earlier quoted context omitted.

HN takes - when fully cached, where it only takes 1 request - ~200ms to load, so 3/4ths of the pageload is just the roundtrip. HN user experience in EU is "meh" because HN is one of the fastest loading dynamic pages on the web to load, meant to give you whiplash from the load speed. Not all pages can load with 1 request, even when designed well, so they will be sub-"meh". HN, if hosted in EU, would have gotten a "daa…

> HN user experience in EU is "meh" It's not: it's super quick, the fastest website I check on a daily basis, by very far. If only all the web could be like this! Which is user tekmol 's point: "Hacker News is one of the most responsive websites I know. And it is run on a single server somewhere in the USA. While I am in Europe." HN is great in the EU. If most of the pageloading is just the roundtrip, you've already…

I disagree, but that does not matter - even if you think it is fast somewhere in the EU, the fact still holds that your perceived loading time is in the ballpark of 2-3x longer than it would have been if hosted within the EU.

How you grade those experiences is subjective. Your "great" is likely my "meh", as physics dictate that we see similar results.