[daper]

CDNPerf and DNSPerf services most likely do testing from some pool of VPSes or hosting services, although I can't see such details in their description. If you are interested how your service performs when serving to other servers, you can use it. If you are interested how it performs when serving to real people (home, mobile, offices), I wouldn't trust it.

Your comparison is missing Akamai which is probably the most feature-rich and largest CDN.

[jakejarvis]

I believe they used to benchmark via their own pool of test servers, but according to the sidebar they're now collecting "real user metrics" somehow:

> All results are based on RUM (Real User Metrics) data from users all over the world

> We gather and analyze more than 300million tests every day

> "RUM Uptime" shows the uptime of CDN providers as measured by real users. Due to bad Internet connections this number can contain false positives.

Agreed about Akamai, though — it appears to be clearly faster than those listed (they certainly had a HUGE head start). But also by far the most expensive. You get what you pay for in this industry!

[krn]

I see CDNPerf and DNSPerf as a way to compare the providers relatively to each other, not as a trustworthy tool go get the real indications of time.

I skipped Akamai, because it seems to be from the period of time, where the number of points of presence (PoPs) was used as the best indicator of the quality of a CDN. It might still be true in some of the third-world countries, but it's no longer the case in the developed world[1].

As for DNS providers: NS1 is like Fastly (the most advanced / premium), Rage4 is like KeyCDN (the best fully-featured value-for-money), and ClouDNS is like BunnyCDN (the best high-volume value-for-money). And Cloudflare is still Cloudflare.

[1] https://www.fastly.com/blog/why-having-more-pops-isnt-always...

[daper]

You're right. I was wondering how this company is able to collect such data from real users. It's registered in an apartment in Cracow, Poland, a 1-person company till last year (here is their company info in polish: http://www.krs-online.com.pl/perfops-sp-z-o-o-krs-10397838.h... , still an apartment address) . Now looking at the Network tab in Developer Tools I can see they are running the tests from JS on their site using visitors of cdnperf.com. I can see requests for URLs like /500b-bench.jpg?t=1552402827417 to various domains.

[jimaek]

CDNPerf is owned by PerfOps, a VC funded startup. Lots of info here https://perfops.net

[ksec]

>NS1 is like Fastly

Except NS1 don't actually offer any pricing.

Rage4, are they being used by any big players site? I mean KeyCDN, despite its price are actually pretty decent and have quite a few Enterprise customers.

And no mention of DNSMadeEasy?

[krn]

> Except NS1 don't actually offer any pricing.

That's true, and it might have been a good enough reason to exclude it from the comparison, as I did with Akamai for CDN. Yet, unlike with Akamai, nothing seems to come really close to NS1 in DNS space.

> Rage4, are they being used by any big players site?

That's its major downside. It has no competition in terms of the price / performance ratio, but might be not big enough company for major players to put a trust in.

Still, Rage4 is regarded with respect by others in its own industry: BunnyCDN uses it for distributing the traffic, and Cloudflare's employee recommends it when Cloudflare doesn't fit[1].

> And no mention of DNSMadeEasy?

I found it to be neither the most technically advanced, nor the best value-for-money. It might be a good option for those who don't have enough justification for NS1, and enough confidence in Rage4.

[1] https://news.ycombinator.com/item?id=12052830

[foobarbazetc]

They’re a good macro view of latency though. You need to do your own testing to get real world results.