MyBB Internal: One or more warnings occured. Please contact your administrator for assistance.
WebPagetest Forums - average connection speeds and impact on load times

WebPagetest Forums

Full Version: average connection speeds and impact on load times
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
I'm trying to figure out how to explain performance gains in layman's terms (to clients).

I'm mostly a front-end/SEO practitioner but I have learned how to enable gzip in .htaccess and some other backend techniques (which is to say I'm not very experienced with networking and traffic).

I have been told that connection speeds which are listed as 'Mbps' are Mega bits per second, and therefore aren't a direct translation to file sizes which are in MB (Megabytes).

How do I translate them and what are 'normal' connection speeds?

For instance, this site lists "Mobile 3G" as an option with a denoted Dowloadspeed of 1.6Mbps, Upload speed of 768Kbps and a Round Trip Time of 300ms. I understand what those things are but how much actual loading time should that (everything else being equal) save if I shave 200 Kilobytes off of the JS.

Even harder to explain, let's say the site loads 60 resources (60 HTTP requests) and I reduce that by 10 (to 50--all from the same domain--I understand how and why this is important). How much time would that generally save?

I don't need anything absolute, just a guideline so when a non-technical person asks (a client or a non-dev co-worker) I can say that 'the time I spent should, as a rule of thumb, speed up the site for a 3G user by 2.5 seconds and a typical broadband user by 1.3 seconds' or whatever.

I just need an approximate conversion between some of these factors and the real world (aside from running before and ofter tests with tools like this).

1 bit = 8 bytes so multiply the file size by 8. Though there's also networking overhead and that assumes perfect utilization. New connections need to warm up first so it's a best-case.

As for estimating impact of changes, the reason you don't see tools providing that already is that it's hard. If the 60 HTTP requests were not on the critical path and were tiny resources then it might not have any impact on the load time at all. If they were all JS files in the head then the impact could be huge. The easiest way to ballpark it would be to use the blocking feature to block those requests and see what the impact is (though depending on the requests you block, it may end up blocking down-stream requests).
Reference URL's