|MadSci Network: Computer Science|
That's an interesting question, and the truth is there are many factors of traffic volume that affect a sites download performance. As you probably know, a user makes a 1 to 1 connection with the web server, but the web server is also making hundreds or thousands of other 1 to 1 connections simultaneously with other users from all over the world. So for each user to enjoy good downloading, the server and its infrastructure must be tuned and working properly for each download to be optimal. Forgetting the users download speed (dial up, DSL, cable, LAN/T1, etc), let's look at the string of possible events which can change download speed. - Bandwidth to the server. This is finite, a server will have allocated a certain amount of bandwidth, and if the number if connections made to the server and number of requests begins to exceed the capacity of the "pipe," then users can see slower downloads. - Database and related issues. So many sites these days are dynamic, driven off database content for each page and object. If the application server which renders the pages or the database which houses the objects experience anything from a bad disk, to too little RAM, to poor internal networking, or a badly tuned database - this can easily cause problems for downloads. As the users queue up, the server will continually be taking too long for each request. - Web servers hosting images. HTML is very lightweight compared to the volume images can be. If a website is poorly created, or simply too heavy in terms of file sizes and badly optimized images, the server can be spending too much time pushing out the payload of images, thus more and more connections will be open at a given time. Web servers want to fulfill requests, close the connection and move on. More and more open connections raises the risks for download issues at the user level. This is where Akamai and their model have gotten good traction, they alleviate the heavy lifting of static images. Good luck, let me know any follow-up questions, Ben
Try the links in the MadSci Library for more information on Computer Science.