::fibreculture:: At least 56K ?

[Adrian Miles]

At 12:50 PM +1100 26/2/2001, melinda rackham wrote:
>does this also apply to getting data out..  i was appalled first 
>time i tried to look at my work in europe and it took forever to 
>download on a really fat university connection..(um forgive lack of 
>techiness)

this is partly anecdotal. i have stuff on servers in the US that has 
loaded blindingly fast at Melbourne University - considerably faster 
than to my desktop from the server next to my desktop on the RMIT 
LAN. i have looked at my content from my RMIT server in Europe and 
America, performance varies but i have not noticed any great 
difference between my content served from australia versus content 
served in the country i'm in.

in terms of your question, yes. there are lots of factors here, 
particularly if you're delivered things like video or audio. 
Basically my server has 10MB ethernet onto a switched 10/100mb 
network (though my building to the main rmit bit is only 10MB), and 
then whatever RMIT has up to its gateway at Melbourne Uni (that is 
big). This lets the server deliver a lot of content, but this content 
is easily choked, as the follow example shows:

i have a 5MB quicktime clip (for instance)
client A requests it. Client A is at RMIT with 10/100MB into their machine.
My server sends client A all the data and bandwidth they can handle 
(http servers are pretty dumb this way). this means, hypothetically, 
client A gets pertty much all my bandwidth for as long as the 
download takes. my server will accept other connections, but they're 
going to get really crappy bandwidth. so my server, which could 
hypothetically handle a 100 simultaenous requests, is slowed to a 
crawl because it's got one client with a ton of bandwidth (same 
applies to somone with a cable modem.)

the solution, in my example, is to use the qtchokespeed tag in my 
quicktime movie, where i can tell quicktime to only deliver, say, 
30Kb of bandwidth with that movie, so no matter how big your pipe you 
won't 'steal' all my bandwidth. (one of the many reasons i tend to be 
known as a quicktime advocate)

you'd be surprised how many people have quicktime or avi's on there 
servers and don't realise that broadband access (on a LAN or cable) 
can actually really slow down their server's performance for other 
clients.

but to the general question. There are numerous things that affect 
access speed, things like the use and proxy servers, speed of 
servers, bandwidth out, etc, as others have commented on. But in 
general it is probably useful to realise that bandwidth in australia 
is not a Telstra monopoly, and so it's general policy issues rather 
than specific commercial practices of a single company (though there 
are issues given Telstra's role as common carrier).

Also I'd caution about generalisations about how bad Australian 
bandwidth is. We need to discriminate who we're discussing, where, 
and where their bandwidth comes from (who provides it). Bandwidth is 
no longer a single entity controlled by Telstra. There are isps who 
have their own pipes to the US, for instance, there's OPTUS of 
course, the Universities, private corporations might have their own 
bandwidth. Here upstream bandwidth is crucial, it's no good having a 
modem pool of 100 if your isp only has a small connection upstream.

other anecdotal observations about bandwidth: in Norway i'm told 
people tend to download from Australian or US servers as it is faster 
than British (never tested it and think its fanciful, but it's the 
perception). Having a fast server with a lots of network access makes 
a major difference to the speed of content delivered, sounds obvious, 
but its surprising the difference it makes to clients. also the 
faster your computer, the faster the network runs in terms of 
downloads. i assume its something to do with the system architecture 
decoding packets but everytime i've upgraded my hardware in my 
university there has been an immediate improvment in network 
performance from my point of view, even where i'm still using 10MB 
ethernet, and there has been no change in network infrastructure. 
same applies to browsers and their rendering engines, and same 
applies to ftp clients (i've had techs. using ftp client x and it is 
lousy, suggest they use ftp client y and their 'bandwidth' improves 
dramatically - its the crappy client their using, not the bandwidth, 
that has changed). and if you're not very 'techie' but are relying on 
your web browser to do your ftp work, well, that's like digging your 
garden with your kitchen fork. There are better tools that do better 
jobs and the blame lies with the tools.

Also, as others have commented, network speed is subject to numerous 
variables. what impresses me in the US when you use a cable modem on 
the eastern or western seaboard is simply how much infrastructure 
there is behind that modem. My content loads damn fast there, that 
means *big* pipes all the way, not just to the home, much like those 
enourmous US freeways, that is as important for broadband 
access/delivery as cable to your home. Not much point all of us 
having cable if the ISP's connection to the world is only 2MB. And 
remember there is *a lot* of the US where you cannot get cable access

finally, costs of broadband access etc, Mark Armstrong presented a 
paper, i think last year (Julian Thomas at Swinburne told me about 
it) which demonstrated that while most US Australian data traffic is 
in our favour, payment is in the favour of the US. This, apparently, 
happens for most countries in relation to the US and so we, in 
effect, are subsidising US internet infrastructure costs which is one 
of the reasons it is cheaper there.

So, from a policy point of view (got there in the end), this is very 
important and requires the renegotiation of various international 
agreements to more accurately reflect the genuine cost of traffic and 
the direction of traffic. this sort of transparency might then help 
in making costs more transparent here and in reducing the cost of 
access. From a policy point of view I also think it is imperative 
that users have a decent sized backchannel. If we can have a cable 
modem which lets my computer become a small server in its own right, 
with my content publically available, then that sort of distributed 
content publication and delivery once again returns the role of this 
technology to the users, which is the point.

cheers
adrian miles
-- 

lecturer in cinema studies and new media rmit university.
lecturer in new media university of bergen.

hypertext theory engine http://bowerbird.rmit.edu.au:8080/
video blog: vog http://hypertext.rmit.edu.au/vog/