Website Performance Data
by Sam AllenProblem
Gain insight into website performance in a specific field. What relationship is there between the number of HTTP requests and the time it takes a site to load? We want to learn the different factors that influence a site's speed. We must measure the total download size, the number of HTTP requests, and the time in seconds to download.
Solution
Test the performance characteristics of 20 websites obtained by a Google search. My experience with the development community is that many searches are for code explanations and samples. The search words I choose to test are "C# Sort".
Search Google for "C# sort"
With this search, I opened the first twenty pages. My developer site was listed twice on the second page of results. I wanted to include it in my comparison. The web browser I used counted the number of HTTP requests, the total size in bytes and also the total time in seconds for the page to completely finish.
What were the results?
My site Dot Net Perls was clearly in the top two places, with another site possibly being faster. The other 18 sites were slower. One blog was over 1 MB big, and the Microsoft blog page took 96 seconds to load. Another site loaded 136 files.
| Activity | Site URL | Page size (KB) | HTTP requests | Transfer time (s) | Comments |
| 1 | google.com | - | - | - | - |
| 2 | devhood.com | 172.85 | 16 | 2.76 | 1 HTTP error |
| 3 | developerfusion.co.uk | 276.72 | 64 | 4.53 | - |
| 4 | dotnetjohn.com | 255.37 | 25 | 2.25 | - |
| 5 | jdconley.com | 484.6 | 77 | 2.27 | - |
| 6 | interact-sw.co.uk | 59.64 | 4 | 1.35 | - |
| 7 | microsoft.co.il (shair) | 304.71 | 44 | - | 96 seconds |
| 8 | en.csharp-online.net | 265.15 | 30 | 2.73 | - |
| 9 | codeproject.com | 439.06 | 64 | 8.46 | - |
| 10 | notaclue.net | 352.73 | 38 | 2.8 | - |
| 11 | 25hoursaday.com | 254.03 | 30 | 4.9 | - |
| 12 | dotnetperls.com | 70.96 | 8 | 0.628 | Two data URLs (not separate HTTP) |
| 13 | dotnetperls.com | 48.41 | 5 | 0.511 | Two data URLs |
| 14 | developers.net | 384.55 | 25 | 1.99 | No article! |
| 15 | csharp411.com | 332.29 | 33 | 2.66 | - |
| 16 | modelmakertools.com | 175.58 | 20 | 2.03 | - |
| 17 | go4expert.com | 807.83 | 136 | 6.09 | - |
| 18 | protocolsofmatrix.blogspot.com | 1208.00 | 90 | 5.56 | 1 HTTP request failed. |
| 19 | builderau.com.au | 357.21 | 72 | 8.13 | - |
| 20 | forums.asp.net | 395.77 | 64 | 5.73 | - |
| 21 | eggheadcafe.com | 159.06 | 30 | 1.44 | - |
What follows here are the above results in graph form. The sites are indexed by a number in the left column of the above table. In the first graph below, Dot Net Perls is highlighted with purple, and in the bottom graph, Microsoft's blog site is highlighted in red (it required 96 seconds to finish, so I left it off the graph).
Why bother?
My investigation here was prompted by Yahoo Exceptional Performance. The above graphs strongly indicate that there is a relationship between number of HTTP requests and download time. However, there is an (equally?) strong relationship between total file size and download time. What are my final conclusions?
Conclusion
It would be wrong to say that the number of HTTP requests determines performance on broadband. Doubtless, reducing the number of requests is beneficial, but my results show that the total time is determined by the number of requests and total size downloaded. Fewer requests and smaller files will always be faster. Finally, there is a big performance difference between many search engine results.