Connection Speeds and Actual Speeds

[Barney_Gumble]

Hello Internet Folk,

Recently I have been experimenting with two different connections to the internet. One is a 44kbps line connection and the other is a 115kbps GPRS link.

I have noticed that the former line connection is faster at actually accessing the web and downloading pages and files than the latter GPRS connection. There is not much in it but definately the 44kbps is faster than the 115 kbps.

I am quite familiar with the packet switching principles of GPRS and the circuit switching of my 44kbps connection, but I have little knowledge of the encoding or language of HTML.

Can anyone give me an explanation :confused:

Regards

Barney

[Crossedcontrols]

I can't answer your question. But if you are looking for a calibrated (of sorts) bandwidth test go here
http://www.beelinebandwidthtest.com/

Regards
CC

[Barney_Gumble]

Thanks CC,

This calibrated test has added an interesting dimension.

The 44kbps link gave a test result of 128kbps (and 2 smiley faces!)

and

The 115kbps link gave a test result of 528kbps (and 1 sad face!)

Now I am confused????

I need to do some more research me thinks.

Barney

[FormerFlyer]

Have a feelin that it gives you a :( whenever you do a test which is slower than any previous ones you have recorded.

cheers ;)

FF

[The Nr Fairy]

Funnily enough this is what I do in my real life, measuring and trying to anticipate application response time over various networks to ensure customers plan WAN upgrades properly.

There are two bits to the connection - one is the bandwidth, the other is the latency. A high bandwidth link with high latency will slow down an application which needs a low latency link,

Likewise, starve a bandwidth-dependent application, and it will suffer.

Check It's the Latency, Stupid! (http://rescomp.stanford.edu/~cheshire/rants/Latency.html) for a good explanation of latency.

It may well be that GPRS adds overhead to the TCP connection which lowers its efficiency below that of the dial-up line.

[WeatherJinx]

TNF

I just got a rating of 240 kbps using a 56k iMac internal modem over AO-Hell (the shame!) - two smiley faces and cheering appaluse, which made this cynical old Hector suspicious immediately!

I'm wondering, in addition to the latency factor that you sagely pointed out, whether page cacheing also has anything to do with it?

If the page is revisited and is already in the user's cache, then the d/l speed is going to be exagerrated, surely? Or am I completely misuderstanding how the test is applied?

I get the feeling the Beeline test is perhaps a bit, well, gimmicky.

Jx

[The Nr Fairy]

WxJx :

Page caching is a way of speeding up subsequent requests, certainly. Whether it helped in your instance is impossible to know. It may be, but it might be someone else had requested that page earlier and it's still live.

When we're conducting the data capture phase of the work, we ensure we capture data while on a LAN, with no other applications running on the PC. And we tend to capture not just HTTP-based applications but SAP, RDBMS, and any other TCP application which a customer has which they think is crucial to their business, so the caching issue is less relevant there.

As an aside, it's interesting to see what apps customers think are critical. They always include "Email and web browsing" and I tend to say - "Nah, it's not vital".

Of course the major difference between the Internet and private circuits is we can give (and meet, for the most part) SLAs - try asking that of A-O-Hell !