What if the route has high latency (which probably is a symptom of congestion too), like i have being on the other side of the world from my server.botg wrote:If the network is congested, how does using multiple parallel transfers make the congestion go away?
I'm assuming that transferring data via FTP involves the client asking for a packet(s) of data, it then being sent by the server, then the client acknowledges it's got it, then asks for the next bit, and so on... (Please do correct me here, i'd like to know... a quick google didn't come up with the juicy details... is it just TCP?)
If the latency is high, the time spent sitting there waiting for the data to start flowing, which may itself come down nice and quick, and then time waiting for the acknowledgement to get back and the next packet sent...
The proportion of time sitting around idle is higher...
So if we send off for two (or more) packets at once, we land up waiting less time for more date...
You mentioned the small file issue, the need for multi-threaded transfers... Isn't this really the same issue? But on a different scale???
Again, since us clients now have nice big fast internet connections, these issues become very visible. And they're only going to get worse for us poor users.
We can all prove this hypothesis very easily. So easily and reliably in fact, that i'd call it a theory.
If you can't see what we're getting at, then you should do some real-world tests. Connect to some servers on the other side of the world. Or next time you get out of Köln and go somewhere a bit more 'back country', internet wise, take your laptop with you. Please.
Filezilla rocks, in every other way...