Intuitively I think it seems that big storage means big upload bandwidth, but I’m not sure what the reality will be. How often will any chunk on a particular drive be accessed? We don’t know. Think about your own hard drive - how much of that data just sits there unread? How much is read/written each day? Obviously it depends on what you do and the kind of data, which makes it hard to envisage what a typical vault might experience. And of course the rank system means that this will vary a lot according to how machine vault configuration and performance (storage, response time, and availability) translates into rank.
One app I’ve been thinking of writing, because it interests me and is simple, is a farming stats app (so I finally posted it: SAFEapp: Farming Performance Tables). App users would put any info in it they want to share about their hardware setup, type of connection, system availability, current rank, farm rate, and that data would be included in a shared database as a sortable/filterable table. I think it would be very popular! Although I’m guessing MaidSafe might have this kind of thing covered.
The reason it is so hard to estimate what bandwidth will be required is because we don’t know where the sweet spots will be. For instance, data that is in high demand will be cached local to the demand. So data that goes viral might get lots of GETS for a while, and then diminish. Data that is accessed regularly but doesn’t tend to hang around in a cache long enough to shortcut many GETS will probably turn out more lucrative over the long term.
I bet MaidSafe have some estimates - or working ranges - but even they don’t know. Its the same for de-duplication. I wouldn’t bet on anything except availability, storage size right now and power consumption right now. Others are thinking response time and bandwidth but I’m not convinced. I’m going to try a couple of low power Odroid-U3 with a big disk on each, sitting on a cable broadband connection (ADSL unfortunately :-()) as my main setup, plus see what happens with laptop (and if possible phone) on very intermittent connections. The data will be very interesting.