Uploading from University :: Blog

Actually, the university graph should be flat for fourty minutes and then rise at the upload speed. This way, it's not representative, if I get it right. The computation should basically be the same as usual accounting does with fixed and variable costs, and the breakeven point.

Hi; where'd you get "40 minutes" from? Also, I'm confused by "flat" - the Y axis is time.

Also, the X axis is MB, not MB/s

Well, my post was a bit inaccurate. Basically, the amount of data transferred doesn't change for the first 25 minutes in the cycle case. Then it is a steady 8 MB/s at which point it actually gets closer to the amount of data transferred by your home line and at some point exceeds it. The amount of time needed to cycle home does not make a difference for the bandwidth, since the transfer can go on, or is finished, but you don't need to count that. The point is that your just not transferring any data while cycling, which isn't reflected well in the graph. (That would be the fixed cost part then.)

should have time along the X axis

Basically, the x-axis is bandwidth, so as pointed out above it should be MB/s. However, otherwise this all makes perfect sense. My only question would be about occasional network failures on campus - do you have to take some sort of expectation with high bandwidth 99% of the time and 0 for the other 1% ? :)

The X axis is *NOT* bandwidth. "MB" is correct. "MB/s" is not. This graph shows how long it takes to upload "x" MB of data.

I get it. You want to know how long it will take to upload a certain amount of data. Each possible amount of data (measured in MB) is listed along the X axis. Look at the Y value for any X value and you will find out how many minutes it takes to upload that number of MB. This makes perfect sense as we're not able to directly adjust time. It should be on the Y axis. The thing we're measuring here is the difference for each quantity of data. That's what we can change, and that's what decides whether it's worth cycling to the university. If we were plotting the number of roadkill occurrences per month, would be put the carcass number on the X axis and the months on the Y axis? No, because presumably it's the month that determines the number of carcasses, not the other way around. It's the number of MB that determines the duration, not the other way around. And from this totally correct graph I get the impression that cycling to uni becomes worthwhile at about 60MB; anything less than that and it might be worth staying home and uploading from there. It was a totally worthwhile exercise and the resulting graph makes a lot of sense. How have Rupert, Ed and company missed this? Too much time measuring download speeds I'm guessing, burned data rates into their minds and made bandwidth become a less practical, more mystical and magical, all pervasive force. :-P Rupert, sebas: You can work out the effective bandwidth in MB/s (remember that we're looking at net figures, not gross) by dividing the value on the X axis by the value on the Y axis and then dividing by 60. If we plotted that we would lose data because the effective bandwidth for the trip is determined by data quantity, not by the passage of time. If you made the changes those guys suggested, it just wouldn't make sense any more. Maybe you should do it, just to show how little sense it makes.