I'm writing a thesis about css sprites and need to know the of the average HTTP request overhead when establishing a connection to download an image file that is a part of DOM.

We conducted a study on one of our personal websites (the median time was 350ms) but this is server specific and can really screw up the perspective we have on the problem.

Does anyone know a study that measured this for more than one server?

An alternative would be some sources about average latency of serving image file by popular dedicated servers (such as nginx or Amazone S3).

  • The latency of any HTTP requests is limited by the speed of light in fiber optics. Consider that the typical speed of light in fiber optic is 200000km/s, a packet of data circumnavigating the earth (40000km) takes 200ms for a single round trip (need to consider: internet backbones doesn't run in a straight line). Add the server processing time, which is dependent on sever implementation, and packet switching overhead along the route, which is highly dependent on congestion in your neighborhood. Those three constitutes the major factors for latency, the latter two are highly variable.
    – Lie Ryan
    Oct 21 '14 at 11:43
  • @LieRyan I am well aware of the factors constituting the latency, I need industry stats to be used as input variables in an optimization algorithm.
    – JohnnyM
    Oct 23 '14 at 7:25

Try to take a look at http://www.speedtest.net/ - perhaps they don't have precisely what you need, but they will have an average data for your broadband provider, which already accumulates a sample of thousands of people :) with a bit of luck you should be able to dig out ping average for your country as well. (because your "request round trip time" is de facto Ping time).

  • maybe I'm blind but I can't find any statistical data for ping on their page (neither can I find anything on Netindex). If you've seen it maybe you could give me the actual address.
    – JohnnyM
    Jan 31 '13 at 23:51
  • @JohnnyM - Try tools.pingdom.com which allows you to break down where the bottlenecks are and also test multiple times to match trends.
    – ionFish
    Feb 1 '13 at 2:16
  • @ionFish This is not what I'm looking for, I need a survey coducted on many servers, preferable high-end ones not a way to test particular websites.
    – JohnnyM
    Feb 1 '13 at 20:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.