I am looking for a way to do realistic testing of web connections through relatively slow ISPs.
I need a test that will give a realistic approximation of how useable an internet connection will be for "web browsing". I realize that there are many different types of activities that this could include, but I mainly mean interacting through a browser with a large number of consumer internet sites, eg Google, news sites, email sites, some file downloading and uploading.
The 'standard' tests don't seem to cover this very well. For example, I've used several well-known bandwidth tests on several different connections and the results were similar, for example .5MB upload and 3MB download. However, on some of these connections that have similar test results I get perfectly usable (if not fast) web browsing, and on others it is nearly unusable browsing. This suggests that the 'standard bandwidth tests' are not really doing a test that's well-correlated with the kinds of real-world use that many people have for routing web browsing and internet use.
This isn't a problem with the machine being used or software on it (eg browser caches, settings, etc) since these are identical across the different tests and connections.
Here are some web-based services for internet bandwidth testing I've used:
The problem is that these services seem to do tests that are not a good simulation of my web browsing usage. In particular, they mostly look at speeds for downloading or uploading a small number of relatively large files, which is good at giving a measure of max bandwidth. Most internet browsing is about lots of little transfers of data and downloading or uploading a large number of small bits of data, and maintaining ongoing connections. For example, clicking on a webpages various buttons and waiting for responses, using remote connection terminals and so forth. These are probably more limited by latency than max bandwidth. Another aspect that the standard tests don't capture at all is that with long latencies it's essentially impossible to use services like VOIP for voice calling. The result is that (for example) while a hard line connection with low latency and .5MB up/3MB down might product perfectly acceptable results, a satellite or other long-latency connection which shows similar connection speeds/bandwidths may be much less desirable to use.
What are the best tests to simulate realistic use? Are there websites that do this? Is there downloadable software (preferably for Mac, linux or mobile device).
Thanks for your help.