[LINK] America's Internet Disconnect

[rchirgwin at ozemail.com.au]

Adrian Chadd wrote:

>On Fri, Nov 10, 2006, Pilcher, Fred wrote:
>
>  
>
>>You're right of course,  I'm being subjective. But by and large, ordinary surfing over their dialup felt like my ordinary surfing here; downloads were a little slower, but not hugely. If I can get 100Kb/s I consider it exceptional - they seemed to get the full 56Kb/s most of the time, which is certainly as good as or better than I get most of the time. 
>>    
>>
>
>It depends where you're downloading from. Well, ok, it depends on the ISP -
>but the speed of light starts playing a part in perceived speed after a
>few thou miles.
>  
>
I'd put the speed of light at bottom of the effects...

Between a US server and here we have (roughly!):
- lots of router hops (from Sydney to samspade.org = 10 hops).
- several points of "oversubscription" (ie, more bandwidth sold than is 
provisioned). Eg: Customers to DSLAM; DSLAM to ISP; Australian ISP's 
transit connection; and transit connection between US ISP and source 
Website).
- Plus 40 milliseconds (approx) speed of light across the Pacific.

What would I bet on causing slow speed, the combination of 
oversubscription and many router hops, or the speed of light?

Oversubscription on DSLAMs ranges upwards of 20 subscribers per unit 
bandwidth to "who knows" in the cheapie services. Oversubscription on 
the DSLAM backhaul isn't published; nor is transit oversubscription; and 
I don't know typical ratios in the US.

If US users get better performance on similar links, I would put it down 
to cheaper transit access leading to less congestion...

RC

>  
>
>>Broadband was another thing. Cases in point: last night a US friend and I were downloading similar-sized files from the same US server at the same time. It took me an hour and a half - it took her about fifteen minutes. I sometimes VNC in to another friend's PC in the US to help sort out his problems - if I download a 10MB utility on his PC it comes down in a matter of seconds whereas it'd take a few minutes for me.
>>    
>>
>
>Try downloading from a fast Australian mirror.
>
>  
>
>>Of course this is no kind of technical analysis. I know that one swallow does not a summer make and, as I said, I don't expect identical performance. But I do get the distinct feeling that we're being diddled.
>>
>>If I'm wrong and just deluding myself, then I apologise.
>>    
>>
>
>Well, you're not -specifically- wrong, but its an interesting case in point.
>Of course Australia is going to be perceived as being slower when accessing US
>sites over, say, someone in the US. There's technologies available to
>mitigate that (for example, have a look at how Google beautifully use HTTP/1.1
>to get the fastest page load times possible by using HTTP pipelining and
>minimal numbers of concurrent connections, over other sites that oft pull in
>ads and media from a dozen different locations - all of which have to finish
>before the page renders) but, by and large, we're going to suck when accessing
>stuff in the US.
>
>What you should do is compare load times to -local- Australian sites. I'm sure
>you'll find news.com.au loads faster for you than it does for your American friend.
>Or using the Optus mirror is faster for you than your American friend.
>(Random data point: 2mbit telstra connection out in the middle of Australia, so its
>an EXPENSIVE 2mbit connection) was able to pull ISOs off the Optus mirror at
>192kbytes/sec.  The only reason it wasn't higher is that the link was being used by
>others (theoretical max 256kbytes/sec; practical max a bit less.)
>
>So using US sites as a measuring stick is always going to be the problem.
>
>
>
>Adrian
>
>_______________________________________________
>Link mailing list
>Link at mailman.anu.edu.au
>http://mailman.anu.edu.au/mailman/listinfo/link
>
>  
>