I am currently hosting my server in US and using my host’s name servers. However I am planning to switch to DNS Made Easy to not only lower my dependence on the host (in case I need to switch from them), but also sort of hoping that it might just speed up site load times, considering majority of the audience is in India and name servers are in US.
Has anyone found dns service providers speeding up load time?
Do you have Google Analytics on your pages? If so you can goo into the speed reporting and look at histograms for the DNS time for your pages from your actual user base (same applies if you are using the Torbit Insight or Sosta RUM tools).
DNS can have a substantial impact on performance if your DNS server is slow and the domain is configured for really short TTLs (time to live). If you get a reasonable amount of traffic and your TTLs are long (hours) then there’s a good chance that the lookups are already cached in the various ISP DNS servers and your origins will not get hit very frequently.
DNS time mostly impacts the base page where it is completely blocking and the resolve time goes directly to the page load time. Best case from latency alone would be an improvement of ~200ms for the requests that do actually have to go back to your DNS servers.
That said, I’m a huge proponent of separating your DNS and domain registration from your hosting provider. Gives you a lot more flexibility and control and dedicated DNS plans are usually cheap enough that it is not cost-prohibitive.
Conclusion:
The quick test dind’t show great numbers and great DNS speed …
[hr]
@ Pat:
In Google Analytics I can see in /content-site-speed/ a … [quote]The average amount of time (in seconds) spent in DNS lookup for this page.[/quote] … of 0,02 seconds, so 20 ms, which sounds good.
Can I check there the DNS Lookup time of each domain that was needed to load the page?
The google analytics information is just for the base page (that is all that is available as part of the w3c navigation timing). You’re also better off looking at the histograms for distribution and not just the averages (I don’t know if 0’s get removed for example). It’s also helpful to look at the distribution filtered for various geographies to see how it looks in various regions of the world.
When resource timing becomes available then you will be able to get timing information for each resource/domain (assuming some privacy headers are added to allow for cross-domain visibility into the performance information).
I’ve been happy with DNS Made Easy for webpagetest.org as well (I like having a stand-alone, just DNS service) but I haven’t really done a lot of independent testing.
btw, the tests you did were using a DSL profile so there was 50ms of additional last-mile latency being injected. If you want to see the raw DNS performance from various test locations you should probably use the “Native” option in the advanced settings (or you can just subtract 50ms from the results).
Funnily enough, my analytic times have actually shot up, once I started using CDN (btw Rackspace CDN, powered by Akamai isn’t being detected as CDN by webpagetest).
It’s possible for render to happen after onload since the browser tends to do a lazy paint but it’s usually on really fast sites with almost no external content (like google.com).
Thank you.
I just had a look at Google Analytics and in my key regions Germany, Austria and Switzerland the DNS Lookup time is quite fast (0,01/0,02 seconds). For people using my website in Pakistan it is quite slow, but I would guess that they will have this problem with most web pages …