measuring connection speed

[mabelis]

Hi, I'm hoping someone can give me a tip: I'm looking for a program to measure the speed of my internet connection. Of course, there are hundreds of sites where I can take a 'speed test' of some sort, but that's not what I'm looking for since taking a speedtest is far off from my usual browsing habits. (I mean: the police wouldn't be very interested in measuring the maximum speed of my car, now would they?) The kind of I'm thinking of runs on my computer and measures times between request and response and the bytes exchanged; this over an extended period of time (a week, say). Surely something like that must exist? I have trouble finding it, though, because I get thousands of hits when looking for measuring connection speed and the first few hundred are all about 'testing' speed, but I don't want to test, I want to measure.
Thanks for any help,
Mabel

[L7Sqr]

The general approach to this type of problem is to collect packet captures (tcpdump, wireshark) and post process them at a later date.

You also mention 'request/response' behavior. What level? The transport layer? Application layer? There are wildly different behaviors at each.

[nuwan4u]

I think York will do it for you, try to google it.It's free

[mabelis]

Sorry for late reaction; I was (involuntarily) taken out for a while.
Thanks for your help; I fear packet analysers are a bit more low-level then I intended; as far as I can make out it would be a hell of a job to distil the info I'm looking for.
I'm interested in request/response bevaviour at the web-browser level. In fact, I don't know whteher my question is even realistic; I wouldn't mind some help formulating it more precisely. The point is, I am struck by all these advertisements promising connections with ever higher speed, while in actual practice many (short) pages take a realy long time materialising on my screen. Due, no doubt, to servers which can't handle the load, bottlenecks somewhere on the path, or even the OS or some piece of software on my computer deciding it got something really important to do first.
This, I feel, makes the advertised speeds, and speed-tests, unrealistic. I have a 10Mbit/s connection at my home and a (much more expensive) 60 Mbit/s glassfiber connection in my shop - yet I fail to notice any difference whatsoever. Except in some (not all) speed-tests. This gets me thinking that the 'nominal' speed of my connection with my ISP has, at least from a certain speed upward, little to do with my browsing experience.
BTW, I tried to find 'York', but couldn't.
So I'm still in the woods... but thanx anyway.
hans

[L7Sqr]

Honestly, the experience you have at the application layer (the layer you are dealing with) is a compound effect of many lower layers. Google has done a large exploration into this space (load improvements at the HTTP level) - you can look at some of the results here (SPDY Protocol) .
There is no way to improve HTTP if TCP (or network fabric) below you is the constraint as HTTP depends on TCP to do it's job.

I'd look at the SPDY link above and see how much of that correlates to what you are seeing. Take pointers and run with some of the ideas there.